Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
DOCTORATE IN TRANSPORTATION AND INFRASTRUCTURES
END OF THE FIRST YEAR EXAMINATION
PhD Student: Brayan Duwan González Hernández
Cycle: XXXIV
Curriculum: Transportation and Land-Use Planning
Tutor: Prof. Luca Persia
Co-Tutor: Dr. Davide Shingo Usami
SAPIENZA UNIVERSITÀ DI ROMA
Department of Civil, Construction and Environmental Engineering
Academic year 2018/2019
III
TABLE OF CONTENTS
LIST OF FIGURES .......................................................................................................................................... V
LIST OF TABLES ......................................................................................................................................... VII
LIST OF ABBREVIATIONS ......................................................................................................................... IX
1 SECTION A: DOCTORAL RESEARCH ............................................................................................. 1
1.1 ADDITIONAL PRELIMINARY KNOWLEDGE ACQUIRED ......................................................................... 1 1.1.1 Courses, Seminars and conferences attended .............................................................................. 1 1.1.2 Books and software packages exploited ...................................................................................... 1
1.2 BIBLIOGRAPHY COLLECTED RELATED TO THE RESEARCH TOPIC........................................................ 2 1.3 STATUS REPORT OF SCIENTIFIC REFERENCE FRAMEWORK, IN RELATION TO THE PROPOSED
RESEARCH TOPIC .............................................................................................................................................. 5 1.4 IDENTIFICATION OF ONGOING SIMILAR RESEARCH ACTIVITIES AT NATIONAL AND INTERNATIONAL
LEVEL 5 1.5 RESEARCH PROPOSAL ........................................................................................................................ 7
1.5.1 Introduction.................................................................................................................................. 7 1.5.2 Objectives ..................................................................................................................................... 8 1.5.3 Methodology ................................................................................................................................ 9
1.6 PUBLICATIONS ................................................................................................................................... 9 1.6.1 Journal Publications .................................................................................................................... 9 1.6.2 Conference Papers ..................................................................................................................... 10 1.6.3 Technical Reports ...................................................................................................................... 10 1.6.4 Invited Presentations ................................................................................................................. 11 1.6.5 Refereeing .................................................................................................................................. 11
2 SECTION B: COLLABORATION AND SUPPORT ACTIVITIES ................................................ 12
2.1 TEACHING SUPPORT ......................................................................................................................... 12 2.2 TRAINING ACTIVITIES ...................................................................................................................... 12 2.3 COLLABORATION WITH RESEARCH AND PROJECTS .......................................................................... 12
3 ANNEXE - LITERATURE REVIEW .................................................................................................. 14
3.1 ROAD SAFETY ................................................................................................................................. 14 3.1.1 Concurrent factors ..................................................................................................................... 16 3.1.2 Road Safety theories .................................................................................................................. 19 3.1.3 Statistical methods to estimate and assess road safety .............................................................. 22
3.2 ROAD TRAFFIC CRASHES DATA ....................................................................................................... 28 3.2.1 Data definitions and standards .................................................................................................. 29 3.2.2 Data collection and storage process.......................................................................................... 45
3.3 EXPOSURE DATA .............................................................................................................................. 48 3.3.1 Population .................................................................................................................................. 49 3.3.2 Driver population....................................................................................................................... 50 3.3.3 Road length ................................................................................................................................ 50 3.3.4 Vehicle fleet ................................................................................................................................ 51 3.3.5 Vehicle kilometers ...................................................................................................................... 51 3.3.6 Person kilometers....................................................................................................................... 51
3.4 ROAD SAFE PERFORMANCE INDICATORS ........................................................................................ 52 3.4.1 SPIs on drink-driving ................................................................................................................. 53 3.4.2 SPIs on the use of protection systems ........................................................................................ 53 3.4.3 SPIs on vehicles ......................................................................................................................... 53
IV
3.5 ROAD INFRASTRUCTURE SAFETY MANAGEMENT ........................................................................... 54 3.6 ROAD INFRASTRUCTURE SAFETY ASSESSMENT METHODOLOGIES ................................................. 57
3.6.1 Road Infrastructure Assessment in Rural Roads ....................................................................... 58 3.6.2 Road Infrastructure Assessment in Urban Streets ..................................................................... 65
3.7 ROAD SAFETY IN DEVELOPING COUNTRIES: EVIDENCE FROM SAFERAFRICA PROJECT.................... 66 3.7.1 Road safety data collection systems in Africa countries ............................................................ 69
V
LIST OF FIGURES
FIGURE 3-1 ROAD SAFETY MAIN CONCURRENT FACTORS (ADAPTED FROM ELVIK ET AL. 2009). ....................... 17
FIGURE 3-2 INFLUENCE OF A ROAD SAFETY MEASURE (ELVIK, 2004). ............................................................... 19
FIGURE 3-3 CASUAL CHAIN MODEL (EVANS, 1991). ........................................................................................... 20
FIGURE 3-4 ELVIK’S REVISED CASUAL CHAIN MODEL. ....................................................................................... 22
FIGURE 3-5 VARIATION OF THE ESTIMATED BEFORE/AFTER EFFECT DEPENDING ON THE NUMBER OF YEARS
CONSIDERED. ............................................................................................................................................. 24
FIGURE 3-6 CONFIDENCE INTERVALS FOR A GAMMA DISTRIBUTION DEPENDING ON THE NUMBER OF YEARS
CONSIDERED. ............................................................................................................................................. 25
FIGURE 3-7 GRAPHICAL ESTIMATION OF THE EXPECTED ACCIDENT THROUGH THE EBM. ................................. 27
FIGURE 3-8 LIFE CYCLE STAGES OF A ROAD INFRASTRUCTURE (OECD/ITF, 2015) ........................................... 54
FIGURE 3-9 DATA REQUIRED AND PURPOSES ASSOCIATED TO EACH PROCEDURE (OECD/ITF, 2015) ............... 57
FIGURE 3-10 PROPORTION OF POPULATION, ROAD TRAFFIC DEATHS, AND REGISTERED MOTOR VEHICLES BY
COUNTRY INCOME CATEGORY (WHO, 2018) ............................................................................................ 67
FIGURE 3-11 RATES OF ROAD TRAFFIC DEATH PER 100,000 POPULATION BY WHO REGIONS:2013, 2016 (WHO,
2018) ......................................................................................................................................................... 67
FIGURE 3-12 SAFERAFRICA OVERALL CONCEPT (SAFERAFRICA, 2016) ............................................................. 68
FIGURE 3-13 EXISTENCE AND USE OF DATABASES – INFORMATION AT NATIONAL LEVEL .................................. 70
FIGURE 3-14 EXISTENCE OF PROCESS EVALUATION FOR SAFETY INTERVENTIONS ............................................. 72
VII
LIST OF TABLES
TABLE 3-1 CONTEXT OF APPLICATION OF RISM PROCEDURES (OECD/ITF, 2015) ........................................... 56
TABLE 3-2 SUMMARY OF ROAD SAFETY INFRASTRUCTURE ASSESSMENT METHODOLOGIES ............................... 58
TABLE 3-3 SUMMARY OF THE MAIN ATTRIBUTES AFFECTING ROAD SAFETY ON RURAL ROADS ......................... 64
TABLE 3-4 BASIC ASPECTS IN MONITORING AND EVALUATION OF ROAD SAFETY DATA COLLECTION PRACTICES
IN AFRICAN COUNTRIES ............................................................................................................................ 70
IX
LIST OF ABBREVIATIONS
AADT Annual Average Daily Traffic
AfDB African Development Bank
ANRAM The Australian National Risk Assessment Model
ARSAP Africa Road Safety Action Plan
ATC Australian Transport Council
AU African Union
AusRAP Australian Road Assessment Program
BCRs Benefit-Cost Ratios
BIR Branch Index Risk
CMF Crash Modification Factor
CTL Research Center for Transport and Logistics
EAT Efficiency Assessment Tools
EB Empirical Bayes
EC European Commission
EU European Union
EuroRAP European Road Assessment Program
EUROSTAT European Statistical Office
GDP Gross Domestic Product
HES Hospital Episodes Statistics
HMI Human Machine Interface
HRS High Risk Sites
HSM Highway Safety Manual
IC Information Centre
iRAP International Road Assessment Program
IRF International Road Federation
IRR Infrastructure Risk Rating
IRTAD International Road Traffic Accident Database
ITF International Transport Forum
LMICs Low- and Middle-Income Countries
OECD Organization for Economic Co-operation and Development
PDO Property Damage Only
PFI Potential for a Safety Improvement Index
NDCs National Data Coordinators
NHS Information Centre of the National Health Service
NO Network Operation
NSR Network Safety Ranking
NTSA National Transport and Safety Authority
NZTA The New Zealand Transport Agency
PIARC World Road Association
RAPs Road Assessment Programmes
RIA Road Safety Impact Assessment
RCA Road Controlling Authority
RFI Risk Factor Index
X
RISA Road Infrastructure Safety Assessment
RISM Road Infrastructure Safety Management
RPS Road Protection Score
RSA Road Safety Audit
RSI Road Safety Inspections
RSO Road Safety Observatory
RTCs Road Traffic Crashes
SCOTI Standing Council on Transport and Infrastructure
SI Safety Index
SIR Section Index Risk
SPF Safety Performance Function
SPIs Safety Performance Indicators
UNECA United Nations Economic Commission for Africa
UNECE United Nations Economic Commission for Europe
usRAP United States Road Assessment Program
VRUs Vulnerable Road Users
WB The World Bank
WRS World Roads Statistics
WHO World Health Organization
1
1 SECTION A: DOCTORAL RESEARCH
1.1 Additional preliminary knowledge acquired
The additional preliminary knowledge acquired within this first year comprise of the
following:
1.1.1 Courses, Seminars and conferences attended
• PhD Seminar Course cycle 34, for first-year PhD students in Transportation and
Infrastructures. Sapienza Università di Roma. 40 Hr., Rome, Italy. 2018-2019.
• 26th World Road Congress - PIARC, Abu Dhabi, UAE. October 5-10, 2019.
• AIIT 2nd International Congress on Transport Infrastructure and Systems in a
changing world, Rome, Italy. September 23-24, 2019.
• SaferAfrica Final Conference. Innovating Dialogue and Problems Appraisal for a
SaferAfrica. Tunis, Tunisia. September 17-18, 2019.
• Road Traffic Injury Prevention and Control in Low- and Middle-Income
Countries. Online course. The Johns Hopkins International Injury Research Unit, 35
Hr., 2019.
• Road Safety in LATAM: from theory to action. Online course. Inter-American
Development Bank (IDB), 35 Hr., 2019.
• Workshop and Training. Road safety risk assessment tool. Monrovia, Liberia. May
26, 2019.
• Workshop and Training. Road safety risk assessment tool. Maputo, Mozambique.
May 21, 2019.
• Final Conference. Modal choice in a multimodal transport system: Tools to
understand the impact of new technologies, walking and cycling measures on modal
choice and on road safety. Brussels, Belgium. May 8-9 2019.
• 4th SaferAfrica Workshop. Dialogue Platform. Brussels, Belgium. April 2-4, 2019.
• SaferAfrica Meeting. WP3: Fostering dialogue on Road Safety and Traffic
Management; and WP5: Road Safety and Traffic Management capacity reviews.
Brussels, Belgium. February 12-13, 2019.
• Continental Workshop on Transport Policy and the African Road Safety Action
Plan (2011-2020). Addis Ababa, Ethiopia. November 19-23, 2018.
• 3rd SaferAfrica Workshop. Innovating Dialogue and Problems Appraisal for a Safer
Africa. Abidjan, Ivory Coast. November 6-7, 2018. Online participation.
1.1.2 Books and software packages exploited
Books:
• Bliss, T., & Breen, J. (2009). Country guidelines for the conduct of road safety
management capacity reviews and the specification of lead agency reforms,
investment strategies and safe system projects.
2
• Bliss, T., & Breen, J. (2013). “Road Safety Management Capacity Reviews and Safe
System Projects Guidelines” Global Road Safety Facility. Washington, DC.
• Elvik, R., Vaa, T., Hoye, A., & Sorensen, M. (Eds.). (2009). The handbook of road
safety measures. Emerald Group Publishing.
• International Transport Forum. (2008). Towards Zero. Ambitious Road Safety
Targets and the Safe System Approach. OECD. Paris.
• International Transport Forum. (2016). Zero road deaths and serious injuries: Leading
a paradigm shift to a safe system. OECD. Paris.
• Muhlrad, N. (2009). Road safety management systems, a comprehensive diagnosis
method adaptable to low- and middle-income countries. Synthèse INRETS.
• World Health Organization. (2018). Global status report on road safety 2018.
• World Health Organization. (2010). Data systems: a road safety manual for decision-
makers and practitioners.
Software packages
• Safety Manager, for manage data relating to traffic, infrastructure and road traffic
crashes.
• Sfinge, for manage and analyze road accident data.
• TransCAD, for store, display, manage, and analyze transportation data.
1.2 Bibliography collected related to the research topic
1. Abdel-Aty, M. (2003). Analysis of driver injury severity levels at multiple locations
using ordered probit models. Journal of Safety Research, 34(5), 597–603.
2. Adminaite, D., Jost, G., Stipdonk, H., & Ward, H. (2016) Ranking EU progress on
road safety: 10th Road Safety Performance Index Report.
3. Amundsen, A. H., & Bjørnskau, T. (2003). Utrygghet og risikokompensasjon i
transportsystemet. En Kunnskapsoversikt for RISIT-Programmet.
4. Appleton, I. (2009). Road infrastructure safety assessment. In 4th IRTAD Conference
(pp. 193–200). Retrieved from
http://internationaltransportforum.org/irtadpublic/pdf/seoul/6-Appleton.pdf
5. Atalar, D., Talbot, R., & Hill, J. (2012). Traiing Package including training manuals
and draft protocols, Deliverable 2.3 of the EC FP7 project DaCoTA.
6. Austroads. (2014). Australian National Risk Assessment Model AP-R451-14.
7. Bliss, T., & Breen, J. (2012). Meeting the management challenges of the Decade of
Action for Road Safety. IATSS Research, 35(2), 48–55.
8. Brodie, C., Durdin, P., Fleet, J., Minnema, R., & Tate, F. (2013). Urban KiwiRAP :
Road Safety Assessment Programme, 1–9.
9. Cafiso, S., Cava, G., & Montella, A. (2007). Safety Index for Evaluation of Two-
Lane Rural Highways. Transportation Research Record: Journal of the
Transportation Research Board, 2019, 136–145. https://doi.org/10.3141/2019-17
10. Cafiso, S., La Cava, G., & Montella, A. (2011). Safety Inspections as Supporting Tool
for Safety Management of Low-Volume Roads. Transportation Research Record:
Journal of the Transportation Research Board, 2203(1), 116–125.
https://doi.org/10.3141/2203-15
3
11. Ceder, A., & Livneh, M. (1982). Relationships between road accidents and hourly
traffic flow—I: analyses and interpretation. Accident Analysis & Prevention, 14(1),
19–34.
12. Chhanabhai, V., Beer, K., & Johnson, M. (2017). Calibrating Infrastructure Risk
Rating ( IRR ) for Victorian Roads. In Australasian Road Safety Conference (pp. 10–
12).
13. De Pauw, E., Daniels, S., Brijs, T., Wets, G., & Hermans, E. (2013). The magnitude
of the regression to the mean effect in traffic crashes. Transportation Research Board.
14. Demasi, F., Loprencipe, G., & Moretti, L. (2018). Road Safety Analysis of Urban
Roads: Case Study of an Italian Municipality. Safety, 4(4), 58.
https://doi.org/10.3390/safety4040058
15. Elvik, R. (2000). How much do road accidents cost the national economy? Accident
Analysis & Prevention, 32(6), 849–851.
16. Elvik, R. (2004). To what extent can theory account for the findings of road safety
evaluation studies? Accident Analysis & Prevention, 36(5), 841–849.
17. Elvik, R. (2006). Laws of accident causation. Accident Analysis & Prevention, 38(4),
742–747.
18. Elvik, R., Vaa, T., Hoye, A., & Sorensen, M. (2009). The handbook of road safety
measures. Emerald Group Publishing.
19. European Commission. (2018). Annual Accident Report 2018, Directorate General
for Transport. Retrieved from
https://ec.europa.eu/transport/road_safety/sites/roadsafety/files/pdf/statistics/dacota/
asr2015.pdf
20. Eurostat. (2003). Glossary for transport statistics. Document prepared by the Inter-
secretariat Working Group on Transport Statistics, Third Edition.
21. Evans, L. (1991). Traffic safety and the driver. Science Serving Society.
22. Hagstroem, L., Fagerlind, H., Danton, R., Reed, S., Hill, J., Martensen, H., … P., T.
(2010). Report on purpose of in-depth data and the shape of the new EU-
infrastructure, Deliverable 2.1 of the EC FP7 project DaCoTA.
23. Harwood, D. W., Council, F. M., Hauer, E., Hughes, W. E., & Vogt, A. (2000).
Prediction of the expected safety performance of rural two-lane highways. United
States. Federal Highway Administration.
24. Hasmukhrai, U. D., Ganeshbabu, K. V, & Gundaliya, P. J. (2016). Identification of
Crash Risk Index for Urban Road: A Case Study of Ahmedabad City. International
Journal of Innovative Research in Technology, 2(12), 2349–6002.
25. Hauer, E., Harwood, D. W., Council, F. M., & Griffith, M. S. (2002). Estimating
safety by the empirical Bayes method: a tutorial. Transportation Research Record,
1784(1), 126–131.
26. Himes, S. C., Donnell, E. T., & Porter, R. J. (2010). Some New Insights on Design
Consistency Evaluations for Two-lane Highways. In 4th International Symposium on
Highway Geometric DesignPolytechnic University of ValenciaTransportation
Research Board.
27. IRAP. (2009). Star Rating Roads for Safety: IRAP Methodology.
28. Ivan, J. N., Wang, C., & Bernardo, N. R. (2000). Explaining two-lane highway crash
rates using land use and hourly exposure. Accident Analysis & Prevention, 32(6),
787–795.
4
29. Kopits, E., & Cropper, M. (2003). Traffic fatalities and economic growth. The World
Bank.
30. Laureshyn, A., Svensson, Å., & Hydén, C. (2010). Evaluation of traffic safety, based
on micro-level behavioural data: Theoretical framework and first implementation.
Accident Analysis & Prevention, 42(6), 1637–1646.
31. Lord, D., Manar, A., & Vizioli, A. (2005). Modeling crash-flow-density and crash-
flow-V/C ratio relationships for rural and urban freeway segments. Accident Analysis
& Prevention, 37(1), 185–199.
32. Miaou, S.-P., Song, J. J., & Mallick, B. K. (2003). Roadway traffic crash mapping: a
space-time modeling approach. Journal of Transportation and Statistics, 6, 33–58.
33. Milton, J., & Mannering, F. (1998). The relationship among highway geometrics,
traffic-related elements and motor-vehicle accident frequencies. Transportation,
25(4), 395–413.
34. Mohamed Eltayeb Zumrawi, M. (2016). Investigating Risk Factors Influencing
Safety in National Highways in Sudan. American Journal of Civil Engineering, 4(6),
276. https://doi.org/10.11648/j.ajce.20160406.12
35. Montella, A. (2005). Quantitative Safety Assessment Methodology, (1922), 62–72.
36. New Zeland Transport Agency - NZTA. (2013). High-risk intersections guide.
37. OECD/ITF. (2015). Road Infrastructure Safety Management Evaluation.
38. OECD/ITF. (2018). Road Safety annual report 2018.
39. Oh, J., Lyon, C., Washington, S., Persaud, B., & Bared, J. (2003). Validation of
FHWA crash models for rural intersections: Lessons learned. Transportation
Research Record, 1840(1), 41–49.
40. Papadimitriou, E., & Yannis, G. (2013). Is road safety management linked to road
safety performance? Accident; Analysis and Prevention, 59C, 593–603.
41. Peden, M., Scurfield, R., Sleet, D., Mohan, D., Hyder, A. A., Jarawan, E., & Mathers,
C. D. (2004). World report on road traffic injury prevention.
42. Resende, P. T. V, & Benekohal, R. F. (1997). Effects of roadway section length on
accident modeling. In Traffic Congestion and Traffic Safety in the 21st Century:
Challenges, Innovations, and OpportunitiesUrban Transportation Division, ASCE;
Highway Division, ASCE; Federal Highway Administration, USDOT; and National
Highway Traffic Safety Administration, US.
43. Rosolino, V., Teresa, I., Vittorio, A., Carmine, F. D., Antonio, T., Daniele, R., &
Claudio, Z. (2014). Road Safety Performance Assessment: A New Road Network
Risk Index for Info Mobility. Procedia - Social and Behavioral Sciences, 111, 624–
633. https://doi.org/10.1016/j.sbspro.2014.01.096
44. Shankar, V., Mannering, F., & Barfield, W. (1996). Statistical analysis of accident
severity on rural freeways. Accident Analysis & Prevention, 28(3), 391–401.
45. Tate, F. (2015). Urban kiwiRAPand IRR Innovation across New Zealand, iRAP
Innovation Workshop 2015. London.
46. Thomas, P., Welsh, R., Mavromatis, S., Folla, K., Laiou, A., & Yannis, G. (2017).
Deliverable 4.1: Survey results: Road safety data, data collection systems and
definitions. SaferAfrica project.
47. Treat, J. R., Tumbas, N. S., McDonald, S. T., Shinar, D., Hume, R. D., Mayer, R. E.,
… Castellan, N. J. (1979). Tri-level study of the causes of traffic accidents: final
report. Executive summary.
5
48. Wang, C., Quddus, M. A., & Ison, S. (2012). Factors Affecting Road Safety: A Review
and Future Research Direction.
49. World Health Organization. (2011). Data Systems. A road safety manual for decision
makers and practitioners.
50. World Health Organization. (2018). Global status report on road safety 2018.
Retrieved from http://e-journal.uajy.ac.id/14649/1/JURNAL.pdf
51. World Road Association. (2015). Road Safety Manual: A Guide for Practitioners.
Paris.
52. Wu, K.-F., Donnell, E. T., Himes, S. C., & Sasidharan, L. (2013). Exploring the
association between traffic safety and geometric design consistency based on vehicle
speed metrics. Journal of Transportation Engineering, 139(7), 738–748.
1.3 Status report of scientific reference framework, in relation to the
proposed research topic
See detailed literature review attached (ANNEXE)
1.4 Identification of ongoing similar research activities at national and
international level
A number of methodologies mostly based on the physical characteristics of a road have been
proposed over the last 15 years by researchers from around the world, especially from Italy
and New Zealand, so far to assess the safety performance of road infrastructures. Probably,
the most known methodology is the international Road Assessment Program (iRAP).
iRAP is the umbrella organisation for EuroRAP, AusRAP, usRAP, and KiwiRAP.
iRAP is based on four standardised protocols that together provide consistent safety ratings
of roads across borders. Nationally, they enable the identification of the most dangerous
roads, tracking performance over time, and therefore where the action is appropriate.
Internationally, they enable comparisons of risk within and between countries. Standard
protocols for iRAP are:
• Risk Mapping: based on real crash and traffic data, colour-coded maps show a
road's safety performance by measuring and mapping the rate at which people are
killed or seriously injured. Different maps can be produced depending on the target
audience.
• Performance Tracking: identifies whether fewer people are being killed or
seriously injured on individual routes or road networks over time, and importantly,
through consultation with road authorities, identifies the countermeasures that are
most effective.
• Star Rating: using drive-through inspections of routes in specially equipped
vehicles. Ratings show the likelihood of a crash occurring and how well the road
would protect against death or serious injury in the event of a crash.
6
• Safer Roads Investment Plans: Following road inspections and coding, in addition
to detailed reporting, a Safer Roads Investment Plan can be developed, considering
over 70 proven road improvement options.
iRAP consisting of a number of evaluation tools; among them, the most relevant to this
project is the Road Protection Score (RPS). The RPS module assigns a road infrastructure
safety level basing on how effectively the infrastructure prevents crashes and protects users
involved in crashes. Based on the calculated RPS the road section is classified according to
a five-level ranking (Star Rating).
iRAP methodology is the inspection of the road network in order to define the level of
safety inherent the road design: five-star roads (green) are the safest, and one-star (black) are
the least safe. Star Ratings can be completed without reference to detailed accident data,
which is often unavailable in low- and middle-income countries. Using specially equipped
vehicles, software and trained analysts, RAP inspections focus on more than 30 different road
design features that are known to influence the likelihood of a crash and its severity. These
features include intersection design, road cross-section and markings, roadside hazards,
footpaths, and bicycle lanes.
Two types of road inspections are available, drive-through inspections and video-based
inspections, with video-based inspections being the most common.
Drive-through inspections require inspectors to record road design data as they drive
along the road using a specialised data tablet. The process is technical and requires accredited
RAP inspectors. Drive-through inspections are typically used where the length of the road
network being surveyed is short or relatively simple (such as rural roads with no adjacent
development). The drive-through inspection equipment includes a video camera, touch-
sensitive laptop, and a GPS antenna. The inspections are followed by a period of data analysis
and quality checking.
Video-based inspections are undertaken in two stages. Firstly, a specially equipped
survey vehicle records images of the road as it travels along. The video is later viewed by
analysts, or coders, and assessed according to RAP protocols. The survey vehicle can record
digital images of the road (generally at intervals of 5-10 metres) using an array of cameras
aligned to pick up panoramic views of the road (forward, left-side and right-side). The main
forward view is calibrated to allow measurements such as lane width, shoulder width, and
distance to roadside hazards. The vehicles can drive along the road at almost normal speed
while collecting the information.
Following the completion of the video-based inspection, each relevant design feature
is measured and rated according to RAP protocols. The process involves streaming the video
images together to form a video of the road network. Coders then undertake desktop
inspections by conducting a virtual drive-through of the road network, at posted speed or on
a frame-by-frame basis, depending on the complexity of the road. The software used by the
coders enables accurate measurements of elements such as lane widths, shoulder widths, and
distance between the road edge and fixed hazards, such as trees or poles. To support the
process a detailed road inspection manual is available. At the completion of the rating
process, it is possible to produce a detailed condition report of the road that forms the basis
for Star Ratings and the Safer Roads Investment Plan. A colour coded map illustrating the
level of safety inherent the road design and features is produced and can be used to make
drivers aware of the risk of different roads or networks (OECD/ITF, 2015).
7
1.5 Research proposal
Title: Development of Simplified Road Safety Methodology for Infrastructure Risk
Assessment
1.5.1 Introduction
Road safety is one of the most critical problems of human life. In fact, around 1.35 million
people die and 50 million are injured in road crashes every year (World Health Organization,
2018). Road traffic crashes are estimated to be the ninth leading cause of death and
projections reveal that it will be the third leading cause of death by 2020 (Peden et al., 2004).
In addition, 90% of the related deaths resulting from road traffic crashes (RTCs) occur in
Low- and Middle-Income Countries (LMICs) (World Health Organization, 2018). At the
same time, LMICs have not fully established crash databases reducing their ability to identify
and measure road safety problems (World Road Association, 2015). Indeed, the fewer the
accident data, the less the information accidents can give about accidents to be prevented
(Montella, 2005).
The cost associated with deaths and injuries is estimated to be in the range between 1.3
and 3.2% of the GDP per annum for many countries (Elvik, 2000). To this regard, traffic
accident prevention has been a consensus all the time around the World and in the last several
years a large amount of money has been spent on traffic accident prevention. Reduction of
social and economic costs also associated with accidents and collisions in road transportation
(Hasmukhrai, Ganeshbabu, & Gundaliya, 2016).
A road traffic crash results from a combination of several factors, in particular, the
accident risk, in terms of repeatability, localization, and severity, is related to three concurrent
factors: infrastructure, vehicle, and human factors (Elvik, Vaa, Hoye, & Sorensen, 2009). In
this way, road and roadside characteristics are a pivotal factor in the number of fatalities and
serious injuries (Chhanabhai, Beer, & Johnson, 2017).
However, progress has been made by some countries in mitigating the number and
severity of road accidents (Adminaite, 2016), but the situation in most low- and middle-
income countries is alarming and even getting worse (Bliss & Breen, 2012). Efforts are being
made towards ameliorating the situation but the efforts are often non-systematic, fragmented
and not knowledge-based or data-led resulting in unsuccessful actions. Nevertheless,
successful road safety actions need to be conducted within the framework of a functional
road safety management system to yield expected results (Papadimitriou & Yannis, 2013).
Road Infrastructure Safety Management (RISM) refers to a set of procedures that
support a road authority in decision-making related to improving the road safety of a road
network. RISM procedures are effective and efficient tools to help road authorities reduce
the number of accidents and casualties, because design standards alone cannot guarantee road
safety in all conditions. Yet successful implementation of RISM procedures requires an
8
adequate level of investment, supporting regulation, availability of relevant road safety data
and adequate institutional management capacity (OECD/ITF, 2015).
A number of methodologies mostly based on the physical characteristics of a road have
been proposed by road safety research so far to assess the safety performance of road
infrastructures (Appleton, 2009).
Probably, the most known methodology is the international Road Assessment Program
(iRAP) consisting of a number of evaluation tools; among them the most relevant to this
research is the Road Protection Score (RPS). The RPS module assigns a road infrastructure
safety level basing on how effectively the infrastructure prevents crashes and protects users
involved in crashes (iRAP, 2009). Based on the calculated RPS the road section is classified
according to a five-level ranking (Star Rating). The iRAP methodology is complex, it
includes many variables and there are no convincing studies that validate it.
To support the assessment of road safety risks on different roads, the research seeks to
develop and pilot a new simplified methodology to quickly identify critical sections and at
low cost even without sufficient crash database. The simplified methodology developed will
be tested and validated through a pilot road safety assessment of highways in Italy,
Mozambique and Liberia.
1.5.2 Objectives
The general objective of this research consists in developing and piloting a new simplified
methodology for road infrastructures’ safety assessment. The underpinning idea is to be able
to recognize road safety issues connected with road infrastructure characteristics, rapidly and
without the specific need for road traffic crash data. The simplified methodology developed
will be tested and validated through a pilot road safety assessment of highways in Italy,
Mozambique and Liberia.
To achieve this objective, the following scientific and technical objectives are considered:
• Reviewing of the knowledge from available research on the most important road
attributes, including the impact of the geometry and operational information of the
roads on road safety risk.
• Choose a set of the attributes to be utilized for the simplified methodology,
considering impact on road safety risk, and feasibility of automated image analysis.
• Development of a standard of video filming for data collection and analysis of road
safety.
• Establishing a methodology for simplified road safety assessment based on the
analysis of road infrastructure attributes (i.e. on their contribution to the risk of road
traffic crashes).
• Support the development of a simplified road risk assessment software using an
automated image analysis and coding tool based on the developed methodology.
9
• Conducting a pilot assessment of national highways in Italy, Mozambique and
Liberia.
• Investigating the relationship between the simplified methodology developed and
road traffic crashes data.
1.5.3 Methodology
A number of methodologies mostly based on the physical characteristics of a road have been
proposed over the last 15 years by researchers from around the world, especially from Italy
and New Zealand, so far to assess the safety performance of road infrastructures.
Road inspections are commonly used as method for assessing risks of traffic crashes.
The most known methodology is the international Road Assessment Program (iRAP). The
iRAP has developed methodologies for visual inspection of roads. With the drive-through
inspections a passenger records the road infrastructure elements while traveling, supported
by a software to rapidly mark the elements and eventually by a camera for post-check. As an
alternative, also video-based inspections can be performed with equipped survey vehicle that
records images of a road at intervals of 5-10 meters. In this case, the inspectors can assess
the road elements while making a virtual drive-through of the road.
The first methodology (drive-through) is technical and entails the presence of trained
inspectors (having iRAP accreditation). On the contrary, the second one (video-based) can
be complex, time-consuming and costly, since it entails the presence of an equipped vehicle
(often to be imported).
The development of the road safety assessment methodology, including the selection
of road infrastructure attributes enabling to assess the road safety risks, will be based on the
extensive international literature available.
The challenge of this study is thus to develop a methodology allowing to overcome the
limitations of the previously mentioned methodologies: being more rapid and less costly than
usual road safety inspections. The research seeks to pilot a new simplified methodology to
quickly identify critical road sections and at low cost, even without sufficient traffic crash
database.
See detailed literature review attached (ANNEXE)
1.6 Publications
Below are the publications, in the framework of the research work developed within this first
year:
1.6.1 Journal Publications
1. González-Hernández, B., Usami, D. S., Prasolenko, O., Burko, D., Galkin, A.,
Lobashov, O., & Persia, L. (2019). The driver’s visual perception research to analyze
pedestrian safety at twilight. Transportation Procedia. [Accepted]
10
2. González-Hernández, B., Llopis-Castelló, D., & García, A. (2019). Operating speed
models for heavy vehicles on tangents of two-lane rural roads. Advances in
Transportation Studies. [Accepted]
3. González-Hernández, B., Usami, D.S. & Persia, L. State of the art of Road
Infrastructure Safety Assessment. [on going]
1.6.2 Conference Papers
1. González-Hernández, B., Meta, E., Persia, L., Usami, D.S., & Cardoso, J.
Identifying barriers to the potential implementation of road safety good practices in
Africa. 99th Annual Meeting of Transportation Research Board, Washington, DC.
January 12-16, 2020. [Accepted]
2. Usami, D.S., Kunsoa, N. B., Persia, L., González-Hernández, B., Meta, E., Saporito,
M. R., Schermers, G., Carnis, L., Yerpez, J., Bouhamed, N., Cardoso, J., Kluppels,
L., & Vandemeulebroek, F. Developing Safe System Projects in Africa. 26th World
Road Congress - PIARC, Abu Dhabi, UAE. October 5-10, 2019.
3. González-Hernández, B., Llopis-Castelló, D. & García, A. Operating Speed models
for heavy vehicles on tangents of Spanish two-lane rural roads. 98th Annual Meeting
of Transportation Research Board, Washington, DC. January 13-17, 2019.
1.6.3 Technical Reports
2019
1. Usami, D. S., & González-Hernández, B. Deliverable 8.15: Report about
Crowdsourcing on road safety in Africa. SaferAfrica project.
2. Deliverable 7.8: Identification of potential local projects. Welsh, R., Kourantidis, K.,
Cardoso, J., Meta, E., & González-Hernández, B. SaferAfrica project.
3. Meta, E., Usami, D. S., González-Hernández, B., Kluppels, L., Viera-Gomes, S.,
Nkeng, G. E., & Wounba, F. Deliverable 6.5: Report on twinning program in
Cameroon. SaferAfrica project.
4. Fava, A. & González-Hernández, B. Deliverable 2.7: Network expansion report 2.
SaferAfrica project.
5. Goldenbeld, C., Kluppels, L., Carnis, L., Cardoso, J., González-Hernández, B.,
Mignot, D., Usami, D. S., & Schermers, G. Deliverable 3.3: Road Safety and Traffic
Management Initiatives. SaferAfrica project.
6. Fava, A. & González-Hernández, B. Deliverable 2.5: Activities Report 4.
SaferAfrica project.
11
7. Tripodi, A., González-Hernández, B. & Shevchenko, A. Final Report. Development
of a new simplified methodology for road infrastructures’ safety assessment based on
the automated analysis of video images.
8. Goldenbeld, C., Carnis, L., Kluppels, L., Usami, D. S., González-Hernández, B., &
Schermers, G. Deliverable 3.2: Road Safety Policy Initiatives. SaferAfrica project.
9. González-Hernández, B. Deliverable 8.13: Report about Crowdsourcing -
SaferAfrica Webinars.
10. Meta, E., González-Hernández, B., Cardoso, J. & Welsh, R. Deliverable 7.2:
Transferability Audit. SaferAfrica project.
2018
1. Usami, D.S. & González-Hernández, B. Deliverable 2.3: Activities Report 2.
SaferAfrica project.
1.6.4 Invited Presentations
1. Simplified Road Safety Methodology for Infrastructure Risk Assessment. 2nd Annual
Training Seminar on SmaLog Issues, Lviv, Ukraine. July 30-31, 2019.
2. Results of Risks Assessment on National Highways in Liberia. Workshop and
Training on road safety risk assessment tool, Monrovia, Liberia. May 24, 2019.
3. Results of Risks Assessment on National Highways in Mozambique. Workshop and
Training on road safety risk assessment tool, Maputo, Mozambique. May 21, 2019.
4. Modal Choice in a Multimodal Transport System. Research Center for Transport and
Logistics (CTL) Workshop, Rome, Italy. May 15, 2019.
5. WP6: Capacity building. Task 6.4 Twining project. EU project SaferAfrica
meeting, Brussels, Belgium. May 3-4, 2019.
6. WP3: Fostering dialogue on road safety and traffic management - Task 3.2 and 3.3 in
Central Africa. EU project SaferAfrica meeting, Brussels. Belgium. February 12-13,
2019.
1.6.5 Refereeing
• Reviewer, Transportation Research Board (TRB)/Transportation Research Record
(TRR); Advances in Transportation Studies (ATS); Drive to the Future project’s
deliverables
12
2 SECTION B: COLLABORATION AND SUPPORT
ACTIVITIES
2.1 Teaching support
• Teaching Assistant to Prof. Luca Persia, Road Safety (Graduate), Sapieza Università
di Roma, Italy, Spring 2019. Lectures:
o Exercises 1: Road accident data analysis
o Exercises 2: Association analysis
o Exercises 3: Analysis of a high crash intersection
o Module 6.1: Road Infrastructure Safety Management (RISM)
o Module 6.3: Road Safety Impact Assessment (RIA)
• Teaching Assistant to Prof. Luca Persia, Transport Policies (Graduate), Sapieza
Università di Roma, Italy, Spring 2019. Lectures:
o Module 3.1: Classification of transport policies: Dissemination of
information
o Module 3.1: Classification of transport policies: Infrastructural measures
o Module 3.1: Classification of transport policies: Infrastructure management
o Module 7.5: Road Safety Impact Assessment (RIA)
2.2 Training activities
• Speaker at the Workshop and Training on road safety risk assessment tool. Results
of Risks Assessment on National Highways in Liberia. Monrovia, Liberia. May 26,
2019.
• Speaker at the Workshop and Training on road safety risk assessment tool. Results
of Risks Assessment on National Highways in Mozambique. Maputo, Mozambique.
May 21, 2019.
• Co-tutor for the Twinning Program between Sapienza Università di Roma and
l'École Nationale Supérieure des Travaux Publics (ENSTP), WP6 – Capacity building
and training actions of the SaferAfrica project. Rome, Italy. April 2019.
• Speaker at the Workshop and Training on road safety assessment of black spots and
high accident risk road sections. Spoleto, Italy. March 6, 2019.
2.3 Collaboration with research and projects
• SAFERAFRICA
SaferAfrica project aims at establishing a Dialogue Platform between Africa and
Europe focused on road safety and traffic management issues. It will represent a high-
13
level body with the main objective of providing recommendations to update the
African Road Safety Action Plan and the African Road Safety Charter, as well as
fostering the adoption of specific initiatives, properly funded.
Funding entity: The European Commission
Duration: Oct. 2016-Sep. 2019
Position: Researcher
• Road Safety Assessment in Mozambique & Liberia
The objective of the project is the development of a new simplified methodology for
road infrastructures’ safety assessment based on the automated analysis of video
images. The simplified methodology being developed will be tested and validated
through a pilot road safety assessment of highways in Mozambique and Liberia.
Funding entity: The World Bank
Duration: Jan. 2018-Jul. 2019
Position: Researcher
14
3 ANNEXE - Literature review
3.1 Road Safety
Road safety is one of the most critical problems of human life. In fact, around 1.35 million
people die and 50 million are injured in road crashes every year (World Health Organization,
2018). Road traffic crashes are estimated to be the ninth leading cause of death and
projections reveal that it will be the third leading cause of death by 2020 (Peden et al., 2004).
Road traffic crashes (RTCs) in the Member States of the European Union claim about
25.600 lives and leave more than 1,4 million people injured in 2016 (European Commission,
2018). In addition, 90% of the related deaths resulting from road traffic crashes (RTCs) occur
in Low- and Middle-Income Countries (LMICs) (World Health Organization, 2018). At the
same time, LMIC's have not fully established crash databases reducing their ability to identify
and measure road safety problems (World Road Association, 2015). Indeed, the fewer the
accident data, the less the information accidents can give about accidents to be prevented
(Montella, 2005).
Besides the human live cost, economic consequences are also very important. The cost
associated with deaths and injuries is estimated to be in the range between 1.3 and 3.2% of
the Gross Domestic Product (GDP) per annum for many countries (Elvik, 2000). The socio-
economic costs of road crashes for the European Union are estimated at least above EUR 500
billion 3% of the EU’s GDP. Most of these costs are related to serious injuries (OECD/ITF,
2018). To this regard, traffic accident prevention has been a consensus all the time around
the World and in the last several years a large amount of money has been spent on traffic
accident prevention. Reduction of social and economic costs also associated with accidents
and collisions in road transportation (Hasmukhrai, Ganeshbabu, & Gundaliya, 2016).
Kopits & Cropper (2003) observed an inverse U-shaped relationship between the capita
GDP and road fatality. Thus, road fatality firstly increases as the economy of a country does,
and therefore decreases when the country becomes developed. The initial growth may be due
to the rapid mobility increase of the country, not in accordance to the road safety knowledge
development. This is typical for developing countries. Developed countries have better
vehicles, infrastructure, knowledge and higher mobility, so the road safety rate decreases
again. This problem reveals as very important if we consider that the number of developing
countries is about to increase during the incoming years.
An accident is defined as an unforeseeable event that alters normal behavior of things
and causes some damage. Thus, a road accident can be defined as an accident in which a
moving vehicle is implied and takes place in the public road network. Accidents are not
completely random. Thus, it is necessary to know and understand their causes, circumstances
and consequences in order to be able to prevent them or, at least, reduce their severity.
Accidents can be classified considering several factors, but the most common are
severity and typology.
According to the damage caused to the people implied in a road accident, victims can
be classified as:
• Fatality. Person who dies instantly or within 30 days after the road accident
takes place.
15
• Injury victim. Person who has been injured as a result of the road accident, but
not resulting in a fatality. We distinguish two types:
o Severe injury. Injury victim who needs to be hospitalized more than 24
h due to the road accident.
o Slight injury. Victim who needs to be hospitalized less than 24 h.
The severity of a road accident is determined as the highest severity level of the people
implied. Therefore, road accidents can be classified as:
• Accident with victims. Accident with at least one victim.
• Fatal accident. Accident with at least one fatality.
• Property Damage Only Accident. Accident with no victims.
The severity of an accident is influenced by several factors, such as the type or road
users, the collision angle and the speed of the vehicles (Laureshyn, Svensson, & Hydén,
2010).
Road accidents can also be classified according to their typology:
• Run off the road accident. The vehicle abandons the platform. The severity of
the accident is highly dependent on the roadside configuration. This is normally
a single-vehicle accident.
• Rear end accident. At least two vehicles are involved, depending this number
on the traffic conditions. The vehicles drive in the same direction and collide
because of the speed dispersion. This accident is very frequent in low-light
conditions, traffic congestion or sudden speed reduction of the preceding
vehicle.
• Head-on accident. Two vehicles driving in opposite directions collide. The
cause of the accident might be diverse. The severity of this accident is normally
maximum, due to the relative speed difference.
• Lateral accident. This accident normally takes place at intersections or curves.
Two vehicles who drive in different (not opposite) directions collide. Its
severity will be determined by the energy dissipated in the collision, as well as
the vehicles type and location of the impact.
A collision implies a sudden kinetic energy release, causing a deformation of the
vehicle(s). Kinetic energy (𝐸𝑘) is determined, depending on the mass (𝑚) of the object and
its speed (𝑣), according to Equation (1).
𝐸𝑘 =1
2∙ 𝑚 ∙ 𝑣2
(1)
Rear-end collisions usually present low severity, since the relative speed differential is
low. On the other hand, head on accidents present the highest relative speed difference, and
therefore the highest severity.
16
3.1.1 Concurrent factors
A road traffic crash results from a combination of several factors, in particular, the
accident risk, in terms of repeatability, localization, and severity, is related to three concurrent
factors: infrastructure, vehicle, and human factors (Elvik, Vaa, Hoye, & Sorensen, 2009).
Other researchers distinguish other two of less importance factors: Traffic and
environmental.
• Infrastructure factor. This factor is related to road design. Road infrastructure
must be designed according to drivers’ expectations. The zones that not meet
the aforementioned condition might present higher crash rates. Some
researchers estimate that this factor is behind over 30% of road accidents, on its
own or combined with human factor. Hence the importance of its consideration
and correct treatment (Treat et al., 1979).
• Human factor. This is the most important concurrent factor, since it is estimated
to be behind over 90% of all road accidents. This factor focuses on the human
being, analyzing both its physical and psychical aspects while performing the
driving task. Its interaction with the infrastructure factor reveals as very
important too.
• Vehicle factor. It focuses on how the vehicle can be involved in the generation
of an accident. It gathers all possible issues with vehicle malfunctions, low
maintenance issues, etc. As the technology develops, this factor reveals as less
important.
• Traffic factor. This is a less important factor than the previous three. Traffic
conditions do also have an effect on road crashes. One example is how the
accident type changes depending on the different traffic states (congested or
free-flow conditions).
• Environmental factor. This is not an important concurrent factor too. It includes
all external factors that may affect the likelihood of having an accident. One
example is weather conditions.
Depending on the factors involved in a road accident, very different solutions may
arise. For instance, some problems related to human factor like drunk driving can be treated
with psychological actions. On the other hand, consistency-related issues should be
addressed through a road redesign. Industrial engineering deals with the vehicle factor. In
addition, in most cases a road accident can be explained through the combination of several
concurrent factors. Hence the importance of multidisciplinary teams to understand road
safety. Figure 3 1 shows the three most important concurrent factors, as well as their relative
importance to road accident likelihood. These factors are also related to the accident severity.
17
Figure 3-1 Road Safety main concurrent factors (adapted from Elvik et al. 2009).
3.1.1.1 Infrastructure factor
Infrastructure plays a major role in accident causation. In fact, this is why accidents tend to
concentrate in certain locations, instead of dispersing randomly through the road network.
Most research focus on the horizontal alignment. Complex alignments are normally
related to higher accident rates. Shankar, Mannering, & Barfield (1996) found that the
increased number of horizontal curves per kilometer increased the severity of the accidents.
Milton & Mannering (1998) found that short road sections were less likely to experience
accidents than longer sections.
Some other researchers found that a higher curvature is linked to a lower accident rate,
which is counter-intuitive (Wang, Quddus, & Ison, 2012). However, this might be because
of the way the curvature was analyzed in that research. The difficulty at analyzing the paper
of the road infrastructure on crashes is that it is normally linked to the human factor. This is
why sometimes road users drive more carefully at more complex alignments.
Some of the most important aspects related to the infrastructure factor are:
• Road type and design-related parameters (design speed, etc.).
• Horizontal alignment.
• Vertical alignment.
• Combined horizontal and vertical alignment, paying special attention to sight distance
and road perception.
• Cross-sectional parameters. Particularly important are the lane and shoulder widths,
since they are highly connected to operating speed.
• Road margins.
• Road marking and signs.
• Pavement conditions.
18
3.1.1.2 Human factor
Human factor considers the issues related to driver reactions and behavior. This factor is
highly related to human psychology, perception, reaction and learning processes. This is a
complex area, so there exist several theories that try to explain them. These theories allow
researchers to detect which level is more likely to be the cause of a road accident, and hence
actuate on it.
Each driver presents different characteristics, abilities and limitations. They are also
influenced by their particular circumstances, which may be related to the environment or not.
Environment conditions affect all drivers at the same level, whereas personal circumstances
obviously not. Some examples of environment-related circumstances are weather conditions,
urban planning, orography, light conditions and more. Some driver-related circumstances are
stress level, fatigue or alcohol consumption.
Hence, all those circumstances result in a high variability of the responses for the same
road layout. This is the reason why the human-road interaction has to be deeply analyzed.
This would allow engineers to design safer roads for everybody, foreseeing drivers’
reactions.
3.1.1.3 Vehicle factor
This factor becomes less and less important in developed countries, due to the technological
development of vehicles. In fact, vehicle related accidents are mostly due to a poor
maintenance, punctures, blowouts, etc. Nevertheless, it remains as a very important
contributing factor in developing countries, since passive and active safety measures are not
embedded in their vehicles.
3.1.1.4 Traffic factor
Accidents occur when traffic moves. These traffic characteristics affect road safety through
both engineering and behavioral effects. We can distinguish four traffic related parameters:
speed, traffic flow, density and congestion (Wang et al., 2012).
It seems clear that the speed has an influence on road safety. A higher speed implies
more kinetic energy, more distance travelled during the perception and reaction time, and a
narrower vision field. The higher kinetic energy implies a higher severity once the accident
has occurred. However, it is not clear how the speed affects the probability of having an
accident.
The extreme variability between operating speed and crash rates can be explained
through the driver-road interaction. From a physical point of view, a higher speed is linked
to a higher accident risk: there is less time to react, the vision field is reduced, and maneuvers
take more distance to be completed. However, the human factor compensates this, increasing
the attention level and the workload demand. They also are more aware of the surrounding
traffic and leave more distance from the preceding vehicle. The infrastructure effect is not
negligible: the roads with higher design standards are normally those which present higher
speeds.
Although it is not clear whether the average operating speed plays an important role on
the generation of road accidents, it seems clearer that the operating speed dispersion does. A
higher operating speed dispersion implies more interactions between vehicles, increasing the
probability of having a crash.
19
Traffic volume is also related to accidents, especially to accident type. As it will be
later indicated, exposure plays a major role in accident estimation. Ceder & Livneh (1982)
analyzed crash rates for different traffic conditions and found that single and multiple crash
rates behaved in different ways according to the traffic conditions.
Himes, Donnell, & Porter (2010) examined the influence of the hourly traffic volume
on the mean speed and its dispersion. They examined 79 sites of 8 roads in Pennsylvania and
Virginia, finding that the hourly traffic volume was strongly correlated to the speed
dispersion. An increase of 100 vph is associated with a decrease in speed deviation by 1.2
mph. Therefore, a higher traffic volume was found to produce a more uniform flow.
The effect of traffic density on road safety still remains almost unknown. The reason
can be the difficulty of accurately estimating traffic density. Ivan, Wang, & Bernardo (2000)
noticed that single-vehicle accident rate increased as the ratio volume/capacity did, following
a negative binomial distribution. The accident rate was the highest at a low volume/capacity
ratio.
The proportion of heavy traffic also affects crash rates. One of the underlying reasons
is the higher speed dispersion, as well as the more amount of passing maneuvers, being a
higher conflict exposure to head-on crashes.
3.1.1.5 Environment factor
The environment factor covers some other aspects not considered previously, such as weather
conditions, urban planning development, orography, etc. The affection is mostly due to an
impairment by drivers (for instance, sun glares or low visibility).
Shankar et al. (1996) found that rain may increase the possibility of injury rear-end
crashes, if compared with PDO crashes. Abdel-Aty (2003) found that darker periods often
lead to a higher accident severity.
3.1.2 Road Safety theories
Road safety theories try to determine why an accident has occurred. The better knowledge
about the underlying phenomena would let researchers and practitioners to develop more
suitable methods and policies for improving safety.
Figure 3-2 represents the most basic approach to understand how a road safety measure
influences the final outcome of road accidents. A certain road safety measure affects several
risk factors, producing a change in the final outcome, in terms of number of accidents or their
severity.
Figure 3-2 Influence of a road safety measure (Elvik, 2004).
This simple model presents three important problems:
• The number of risk factors that should be considered is very large. Some of them
remain even unknown or unmeasurable.
• Many of the road safety evaluation studies do not clearly identify and/or measure the
risk factors influenced by the countermeasure.
20
• Some road safety measures present user behavioral adaptation, i.e., users get adapted
to the countermeasure by changing their attitudes and behavior. Thus, the safety
measure could indeed be counter-productive.
Evans (1991) suggested a two casual chain model that includes this phenomenon (Figure
3-3).
Figure 3-3 Casual chain model (Evans, 1991).
This duality is the reason why road safety lacks of a solid theoretical ground, on the
contrary to several other mature disciplines (Wang et al., 2012). Instead, there exist some
groups of theories that try to explain the user-road-crashes interaction. We can distinguish
two ways of approaching to road safety:
• By means of the infrastructure factor. Several objective relationships can be
established between some geometric or environmental parameters and road crashes.
• Analysis of the human factor. This approach cannot estimate the number of road
accidents. Instead, a better knowledge of the process is achieved.
There are some other theories that try to combine the best part of both approaches.
Some of them try to explain driver’s attitudes and behavioral change after a certain
countermeasure is applied. Some others establish a general framework for driver behavioral
adaptation due to infrastructure changes.
Elvik (2004) proposed a conceptual framework based on Evans’ model (Figure 3-3).
He proposed the following risk factors to be considered, as well as the behavioral adaptation:
• Kinetic energy. This is not a risk factor per se, since it does not cause harm as long
as it is controlled. If a collision takes place, this energy is released, affecting the
severity.
• Friction. This factor is related to the control and stability of the vehicle.
• Visibility. The more sight distance, the more time drivers have to process the
information, hence reducing the likelihood of surprises.
• Compatibility. It refers to the difference that exists between different types of vehicles
in terms of speed, mass, performance, etc.
• Complexity. It refers to the amount of information that a user has to process per unit
of time.
21
• Predictability. It denotes the reliability at which the occurrence of a risk factor can be
predicted in a given situation.
• Individual rationality. Individual users normally try to behave looking for their
maximum benefit, i.e., satisfying their preferences.
• Individual vulnerability. When an accident occurs, some individuals are more
exposed than others.
• System forgiveness. Some elements of the road should be designed in order to prevent
accidents or reduce their severity. Some examples are clear margins, rumble strips,
road lighting, and others.
In order to prevent counterproductive responses, Amundsen & Bjørnskau (2003) suggested
to analyze the following factors, which already include the behavioral adaptation effect:
• How easily a certain countermeasure is noticed. Drivers are continuously scanning
the road. When they notice a safety countermeasure, behavioral adaptation might
occur. Thus, the best solution is to act without leaving them to know (obviously, this
is not always possible).
• Historical antecedent of behavioral adaptation to basic risk factors. There is a higher
probability of behavioral adaptation if it already took place before.
• Size of the engineering effect on generic risk factors. Large changes are more likely
to be noticed by users.
• Whether or not a measure primarily reduces injury severity. Measures that reduce
injury severity are less likely to lead to behavioral adaptation than measures that
mostly act on reducing the likelihood of an accident.
• The likely size of the material damage incurred in an accident. Road users prefer the
material damage in an accident to be as small as possible.
• Whether or not additional utility can be gained. Users try to maximize utility of the
trip. For some road safety measures, it is difficult to see how road users could gain
any benefit by changing their behavior.
Considering all these parameters, Elvik (2006) proposed a revised causal chain model that
incorporated the relationships between road safety measures and driver behavior, through
behavioral adaptation (Figure 3-4). The result is termed as behavioral safety margin,
indicating how road users assess their safety margin when travelling.
22
Figure 3-4 Elvik’s revised casual chain model.
According to Elvik (2006), accidents may be explained according to a few general statistical
regularities that determine the relationship between risk factors and accident occurrence.
These regularities are called “laws of accident causation”. He proposed the following laws:
• Universal law of learning. The ability to foresee undesirable traffic situations
increases uniformly as the amount of travel (or conflicts) increases. This law also
implies that the accident rate per unit of exposure decreases as the exposure increases.
• The law of rare events. The rarer a certain risk factor is encountered, the larger its
effect results on accident rate. Moreover, its rareness makes this event more difficult
to be learnt.
• The law of complexity. The more information rates the road user must attend to, the
higher the probability of committing an error.
• The law of cognitive capacity. As the cognitive capacity of a road user approaches to
their limits, the higher the probability of having an accident.
3.1.3 Statistical methods to estimate and assess road safety
There exist some specific tools for estimating or analyzing crashes. Some of them allow the
designers to estimate the number of accidents depending on some factors. Some others are
useful for determining whether a road countermeasure has been effective or not.
3.1.3.1 Safety Performance Functions
A Safety Performance Function (SPF) is an expression that allows us to estimate the number
of crashes in a certain roadway entity depending on some factors. The factors include some
design and/or environmental features, as well as the exposure. The exposure may have an
influence on the output or not. Those functions are normally calibrated considering a
Negative Binomial distribution.
Their common functional form is shown in Equation (2) (intersections) and Equation
(3) (road segments). AADT and length are normally given in vpd and km, respectively.
𝜆𝑖 = 𝐸(𝑦𝑖) = 𝛽0 ∙ 𝐴𝐴𝐷𝑇𝑖𝛽1 ∙ 𝑒∑ 𝛽𝑗∙𝑋𝑖𝑗
𝑘𝑗=2 (2)
23
𝜆𝑖 = 𝐸(𝑦𝑖) = 𝛽0 ∙ 𝐴𝐴𝐷𝑇𝑖𝛽1 ∙ 𝐿𝑖
𝛽2 ∙ 𝑒∑ 𝛽𝑗∙𝑋𝑖𝑗𝑘𝑗=2 (3)
𝑋𝑖𝑗 represents the different parameters that are considered by the SPF, while 𝛽𝑖𝑗 are the
corresponding estimates. The exposure is normally introduced in terms of elasticity. This is
the functional form that produces the best adjustments (Oh et al., 2003).
The exposure is very important in those models. In fact, it explains most of the accident
variability. However, the way to consider it has been very controversial. Some researchers
support that the exposure does not affect the crash generation process, and so assuming 𝛽1 = 𝛽2 = 1. In recent years, most researchers assume that the AADT has a true effect on how
accidents are generating, thus not enforcing 𝛽1 = 1.
According to the AADT estimate, there are four possibilities:
• 𝛽1 = 0. The number of crashes is not influenced by the traffic volume. Obviously,
this is not true.
• 𝛽1 = 1. The crash rate is the same regardless of the traffic volume. The number of
crashes is proportional to AADT.
• 𝛽1 > 1. The crash rate becomes higher as the traffic increases.
• 𝛽1 < 1. The crash rate becomes lower as the traffic volume increases. This is the most
common outcome for the AADT estimate, according to most safety performance
functions.
The consideration of the segment length has remained more controversial. Some
researchers include it in the analysis, obtaining a calibrated estimate. Some others do not,
fixing it to 1 but performing a negative binomial regression, which may also be correct. In
the last case, researchers assume that the road segment length does not have an influence on
the crash rate. Some researchers indicate that it behaves in the opposite direction than AADT:
a longer road segment leads to a higher crash rate. Some others, like Miaou, Song, & Mallick
(2003) and Lord, Manar, & Vizioli (2005) affirm that road length does not affect crash rates.
Obviously, the length of the road segment might only be relevant if homogeneous road
segments are considered. Thus, road segmentation becomes a very important issue. (Resende
& Benekohal, 1997) indicated that only homogeneous road segments should be considered,
based on traffic flow and geometric characteristics.
3.1.3.2 Before/After studies
Before/After studies are widely considered to be the most appropriate method to execute the
evaluation of the effectiveness of traffic safety measures (De Pauw, Daniels, Brijs, Wets, &
Hermans, 2013). It consists on comparing the number of accidents before and after the
application of the countermeasure.
Although this may seem a simple approach, there are some problems due to the nature
of road accidents. De Pauw et al. (2013) distinguished the following issues:
• Regression to the mean.
• Long-term trends affecting the number of crashes or injured road users.
• General changes in the number of crashes.
• Changes in traffic volumes.
• Any other specific events introduced at the same time as the road safety measure.
24
Due to the high variability of road crashes, the actual number of accidents at a certain
location can never be known. However, the more years of data we have, the more precision
about the outcome. When comparing the number of accidents before and after a
countermeasure has been applied, at least 3-5 years before and after are suggested to use.
Figure 3-5 shows how the accident randomness affects the results.
Figure 3-5 Variation of the estimated before/after effect depending on the number of years
considered.
Several researchers have stated that the distribution of the expected mean of a Poisson-
distributed count parameter follows a Gamma distribution. Considering this assumption, we
cannot perfectly estimate the expected number of accidents, but we can determine a range
that includes it with a certain probability.
According to it, we can use the properties of the Gamma distribution to estimate the
range within the actual expected number of accidents is located. Figure 3-6 represents the
variation of the lower and upper bound of the range for an estimation of three crashes/year.
One can notice how the uncertainty is extreme for 1-2 years, but it is quite stable for more
than 5 years. This is why at least 3 to 5 years are recommended to be used for before/after
analyses. This is due to the Regression to the Mean (RTM) bias (De Pauw et al., 2013). If
short periods of time are considered, the Empirical Bayes Method is suggested as a good tool
to reduce this bias. If long periods of time are considered, there is no need to use an additional
technique.
25
Figure 3-6 Confidence intervals for a gamma distribution depending on the number of years
considered.
The accident outcome after the application of the countermeasure can also be affected
by some other factors. Some examples are social awareness, traffic volume variations, etc.
Those factors cannot be directly measured but they do exist. Thus, the effect of those other
factors should be deducted in order to estimate the actual effect of the safety measure. We
can do this by examining the crash variation in a control group. A control group is a set of
similar roadway entities in which the countermeasure has not been applied. Thus, the
variation of the number of crashes is only due to these general factors. Their comparison will
let us to determine the true effect of the countermeasure.
3.1.3.3 Crash Modification Factors
A Crash Modification Factor (CMF) is a coefficient that lets us rapidly estimate the variation
of the crash outcome due to a certain countermeasure. Considering 𝑦0 the initial number of
accidents of the roadway entity 𝑖, the number of accidents after the countermeasure is applied
(𝑦0) can be calculated as shown in Equation (4).
𝑦𝑓 = 𝑦𝑖 ∙ 𝐶𝑀𝐹0→𝑓 (4)
𝐶𝑀𝐹0→𝑓 is the crash modification factor that let us go from the initial to the final
condition. Is worth pointing out that CMFs are normally not considered in terms of before-
after situations, but referred to a base condition. The CMF is 1.0 for the base condition. Some
26
CMFs refer to all accidents, while others refer to a certain subgroup (type of accident or
severity).
Crash modification factors are a very simple and powerful tool, but they have to be
handled with care. They were calibrated based on several Before/After analysis, considering
certain conditions, such as traffic volume, cross-section, visibility, etc. A variation of those
parameters might affect the outcome of crashes. Therefore, CMFs should only be applied
when these additional conditions are satisfied.
There are many situations in which more than one CMF needs to be used. This is not a
problem, as long as all conditions are satisfied. The uncertainty about the outcome also
increases, as further discussed. A general formulation is given in Equation (5) (Wu, Donnell,
Himes, & Sasidharan, 2013). 𝑦𝑟𝑠 is the predicted number of crashes per year on a roadway
element. 𝑦𝑏𝑟 is the predicted number of crashes for the base conditions. 𝐶𝑀𝐹𝑗 are all the crash
modification factors to apply. Finally, 𝐶𝑟 is a calibration factor for the highway element for
local conditions.
𝑦𝑟𝑠 = 𝑦𝑏𝑟 ∙ 𝐶𝑟 ∙ ∏ 𝐶𝑀𝐹𝑗𝑛𝑗=1 (5)
The calibration factor for local conditions covers social, climatic and other aspects that
vary across regions and have a certain effect on the number of accidents.
Sometimes, the CMF is not a single value but a function (Crash Modification
Function). They are basically managed in the same way as crash modification factors.
CMFs are normally calibrated considering several Before/After analyses. Thus, there
exist a certain degree of uncertainty, which is reflected in the variance of the CMF. This
allows us to get an idea about their performance and the validity of the outcome. Of course,
the more CMFs we use in our analysis, the more uncertain the result becomes.
CMFs can be used together with safety performance functions for a better estimation
of the number of crashes, according to the following steps:
1. Estimation of the number of accidents on a road geometric element for the base
conditions. This can be done by means of a safety performance function (𝑦𝑏𝑟).
2. Adjustment of the previous quantity for the local conditions, applying the CMFs and
the geographical parameter (𝐶𝑟). The estimated number of crashes is 𝑦𝑟𝑠.
3. If some information about actual crashes is available, the Empirical Bayes method
can be applied (further explained).
There are tons of crash modification factors available for designers. The AASTHO’s
Highway Safety Manual contains several of them, including their variance, accuracy and
feasibility. All those CMFs covered by the part C of the Highway Safety Manual (HSM)
present a standard error less than 0.1, whereas CMFs that appear on part D present a standard
error lower than 0.3. To identify appropriate CMFs to be applied, a good database can be
found on the web page: www.roadsafety-dss.eu. Finally, CMFs should be handled with care.
No risk exposure is considered, as well as interaction among the different parameters is not
covered.
3.1.3.4 Empirical Bayes Method
The Empirical Bayes Method assumes that accident counts are not the only clue to the safety
of a roadway entity. The other clue is how similar roadway entities behave. For instance, if
we know that a certain roundabout presents 0 accidents in a year, but on average roundabouts
27
present 0.56 accidents in a year, it would not be correct to assume that our roundabout is
completely safe. In the same way, we already know that our roundabout behaves slightly
better than the average roundabout. Hence, the actual crash probability of our roundabout
should be within those values.
According to (Hauer, Harwood, Council, & Griffith, 2002), the Empirical Bayes
Method addresses two safety estimation issues:
• It increases the precision of estimates beyond what is possible when the available data
is limited.
• It corrects the regression to the mean bias.
The Empirical Bayes Method considers both observed and estimated data. The
expected number of accidents is calculated as shown in Equation (6).
𝐸 (𝜆
𝑟) = 𝛼 ∙ 𝜆 + (1 − 𝛼) ∙ 𝑟 (6)
𝐸 (𝜆
𝑟) represents the estimated number of accidents. 𝜆 is the expected number of accidents,
according to the SPF estimation. 𝑟 is the observed number of accidents. 𝛼 is a weight
parameter, that gives more importance to the estimated or the observed accidents, according
to the reliability of the SPF. This parameter is calculated as Equation (7) shows, being 𝜇 the
over dispersion parameter of the SPF.
𝛼 =1
1+𝜆∙𝜇 (7)
Depending on the over dispersion parameter of the safety performance function, the
estimated number of accidents will be closer to the SPF estimation or the observed accidents
(Figure 3-7).
Figure 3-7 Graphical estimation of the expected accident through the EBM.
28
Harwood, Council, Hauer, Hughes, & Vogt (2000) recommend to apply the Empirical
Bayes procedure in the following cases:
• For estimating the number of accidents for the “do-nothing” alternative.
• Projects where the roadway cross-section is changed but the basic number of lanes
remains the same. This includes, for instance, shoulder or lane widening projects.
• Projects with minor changes in the alignment.
• Projects in which a passing lane or a short four-lane section is added to increase
passing opportunities.
• Any combination of the above.
On the contrary, the Empirical Bayes procedure is not applicable in the following cases:
• Projects where there is an important change in the alignment layout.
• Intersections where the number of legs is changed.
3.2 Road Traffic Crashes data
Reliable and consistent road accident data are a valuable and necessary prerequisite for the
support of decision making aimed at the improvement of road safety. Based on the report on
Data Systems (World Health Organization, 2011), some steps are given in order to strengthen
an existing road accident system or design and implement a new one. The basic targets are
considered similar when designing a common data collection system based on the national
existing ones. These steps are the following:
1. Establishing a working group, which will review and discuss the road safety goals set
already by the national lead agency in terms of data requirements for monitoring and
achieving each one.
2. Choosing a course of action, which is a range of strategies aiming to strengthen road
safety systems depending on the different needs and characteristics of each region or
country. The main strategies concern:
• the improvement of data quality and system performance of road accident systems
coming from police data
• the improvement of health facility-based data on road injuries.
• the improvement of the vital registration system and particularly the death
registration system
• the combination of existing data sources in order to obtain more accurate
estimates on the magnitude and effects of road injuries
3. Defining the recommended minimum data elements and definitions, based on specific
selection criteria.
The recommendation for a common accident data collection system consists of a
minimum set of standardized data elements, which allows international comparisons to be
made.
For the development of a common data collection system, a two-step approach is most
commonly recommended:
a. improvement and harmonization of existing data and methods
29
b. collection of new harmonized data
The common dataset composed of minimum data elements (variables) will be a key
tool for ensuring the appropriate data are captured to enable analysis, and for maximizing
consistency and compatibility of data collected across different jurisdictions/ countries.
Uniformity of accident data is especially important when combining sub-national datasets
and for international comparisons.
3.2.1 Data definitions and standards
One of the greatest limitations when examining international comparisons of road accident
figures is the incompatibility of data, which is due to either different collection procedures
or different definitions of the variables and values used.
Concerning road fatalities, the uniform international definition of persons killed in
road accidents is defined as “the persons who died within 30 days from the day of the
accident”. At present this definition is used by a number of developing countries and is
suggested to be adopted by the remaining ones. On that purpose, some countries have to
modify the data collection process and develop appropriate conversion factors, for the
conversion of the number of road accident fatalities prior to the adoption of the common
definition.
On the other hand, definitions of injury severity may present important differences
among countries. Furthermore, the minimum injury for which an accident is recorded by the
Police is different in each country. Especially, the distinction between seriously and slightly
injured persons presents important differences among countries.
One of the main problems of each national road accident data file is that not all injury
accidents are recorded. Underreporting is an issue of general concern in developing countries
and affects the degree to which the statistical output of a country’s data system reveals the
actual situation of road safety. Thus, underreporting delivers a biased database in terms of
fatalities and serious injuries. Road accident databases that link Police and hospital data may
serve as a potential solution to the underreporting issue.
However, additional inaccuracies in reporting the various variables and values
contained in the national road accident data collection form may exist. Such vagueness,
which are inherent to the nature of these variables and values, result from the conditions
under which the primary information is collected by the police officer as well as the way this
information is filled-in later on. Such inaccuracies may also raise due to inadequate training
of the Police force collecting the information.
Moreover, two main sources of data incompatibility can be identified and should be
handled:
• incompatibilities due to missing or incomplete national definitions (e.g. for weather
conditions)
• incompatibilities due to different definitions in different countries (e.g. for road
types).
The establishment of international rules for road accident data variables, values,
structure and definitions has been recommended by several international research projects
and some efforts for harmonizing accident data at international level have already taken place
(e.g. CARE system). The data structure, definitions and formats for the most common
variables in road safety analyses is presented in the following sections.
30
However, it should be noted that when planning the introduction of new variables or
modifying the existing ones, changes to the definitions and values of existing data elements
should be minimized, as these can create problems with the consistency and comparability
of data over time. On the other hand, if definition or data element changes are made, then the
date of change should be clearly noted in official records, allowing for some misclassification
during the transition period.
3.2.1.1 Accident data elements
The accident data elements describe the overall characteristics of the accident.
A1. Accident ID
Definition: The accident identification number is a number which will allow the accident
record to be cross-referenced to road, traffic unit and person records. It consists of three
distinct fields, the country code, the year and the accident number.
Obligation: Mandatory
Data type: Numeric or character string
Comments: This value is usually assigned by the police as they are responsible at the accident
scene. Other systems may reference the incident using this number.
A2. Accident date
Definition: The date (day, month and year), on which the accident occurred.
Obligation: Mandatory
Data type: Numeric (DDMMYYYY)
Comments: If a part of the accident date is unknown, the respective places are filled in with
99 (for day and month). Absence of year should result in an edit check. Important for seasonal
comparisons, time series analyses, management/administration, evaluation and linkage.
A3. Accident time
Definition: The time at which the accident occurred, using the 24 hour-clock format (00.00-
23:59).
Obligation: Mandatory
Data type: Numeric (HH:MM)
Comments: Midnight is defined as 00:00 and represents the beginning of a new day. Variable
allows for analyses of different time periods.
A4. Accident municipality and region
Definition: The municipality and county or equivalent entity in which the accident occurred.
Obligation: Mandatory
Data type: Character string
Comments: Important for analyses of local and regional programmes and critical for linkage
of the accident file to other local/regional data files (hospital, roadway, etc.). Also important
for inter-regional comparisons.
A5. Accident location
Definition: The exact location where the accident occurred. Optimum definition is route
name and GPS/GIS coordinates if there is a linear referencing system (LRS), or other
31
mechanism that can relate geographic coordinates to specific locations in road inventory and
other files. The minimum requirement for documentation of accident location is the street
name, the reference point, the distance from the reference point and direction from the
reference point.
Obligation: Mandatory
Data type: Character string, to support latitude/longitude coordinates, linear referencing
method, or link node system.
Comments: Critical for problem identification, prevention programmes, engineering
evaluations, mapping and linkage purposes.
A6. Accident type
Definition: The accident type is characterized by the first injury or damage-producing event
of the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Accident with a pedestrian: Accident between a vehicle and at least one pedestrian.
2. Accident with a parked vehicle: Accident between a moving vehicle and a parked
vehicle. A vehicle with a driver that is just stopped is not considered as parked.
3. Accident with a fixed obstacle: Accident with a stationary object (i.e. tree, post,
barrier, fence, etc.).
4. Non-fixed obstacle: Accident with a non-fixed object or lost load.
5. Animal: Accident between a moving vehicle and an animal.
6. Single vehicle accident /non-collision: Accident in which only one vehicle is involved
and no object was hit. Includes vehicle leaving the road, vehicle rollover, cyclists
falling etc.
7. Accident with two or more vehicles: Accident Accidents where two or more moving
vehicles are involved.
8. Other accident: Other accident types not described above.
Comments: If the road accident includes more than one event, the first should be recorded,
through this variable. If more than one value is applicable, only the one that corresponds best
to the first event should be selected. Important for understanding accident causation,
identifying accident avoidance countermeasures.
A7. Impact type
Definition: Indicates the manner in which the road motor vehicles involved initially collided
with each other. The variable refers to the first impact of the accident, if that impact was
between two road motor vehicles.
Obligation: Mandatory
Data type: Numeric
Data values:
1. No impact between motor vehicles: There was no impact between road motor
vehicles. Refers to single vehicle accident, collisions with pedestrians, animals or
objects.
32
2. Rear end impact: The front side of the first vehicle collided with the rear side of the
second vehicle.
3. Head on impact: The front sides of both vehicles collided with each other.
4. Angle impact – same direction: Angle impact where the front of the first vehicle
collides with the side of the second vehicle.
5. Angle impact – opposite direction: Angle impact where the front of the first vehicle
collides with the side of the second vehicle.
6. Angle impact – right angle: Angle impact where the front of the first vehicle collides
with the side of the second vehicle.
7. Angle impact – direction not specified: Angle impact where the front of the first
vehicle collides with the side of the second vehicle.
8. Side by side impact – same direction: The vehicles collided side by side while
travelling in the same direction.
9. Side by side impact – opposite direction: The vehicles collided side by side while
travelling in opposite directions.
10. Rear to side impact: The rear end of the first vehicle collided with the side of the
second vehicle.
11. Rear to rear impact: The rear ends of both vehicles collided with each other.
Comments: Useful for identifying structural defects in vehicles.
A8. Weather conditions
Definition: Prevailing atmospheric conditions at the accident location, at the time of the
accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Clear (No hindrance from weather, neither condensation nor intense movement of air.
Clear and cloudy sky included)
2. Rain (heavy or light)
3. Fog, mist or smoke
4. Sleet, hail
5. Severe winds (Presence of winds deemed to have an adverse effect on driving
conditions)
6. Other weather condition
7. Unknown weather condition
Comments: Allows for the identification of the impact of weather conditions on road safety.
Important for engineering evaluations and prevention programmes.
A9. Light conditions
Definition: The level of natural and artificial light at the accident location, at the time of the
accident.
33
Obligation: Mandatory
Data type: Numeric
Data values:
1. Daylight: Natural lighting during daytime.
2. Twilight: Natural lighting during dusk or dawn. Residual category covering cases
where daylight conditions were very poor.
3. Darkness: No natural lighting, no artificial lighting
4. Dark with street lights unlit: Street lights exist at the accident location but are unlit.
5. Dark with street lights lit: Street lights exist at the accident location and are lit.
6. Unknown: Light conditions at time of accident are unknown.
Comments: Information about the presence of lighting is an important element in analysis of
spot location or in network analysis. Additionally, important for determining the effects of
road illumination on night-time accident accidents to guide relevant future measures.
3.2.1.2 Accident data elements derived from collected data
AD1. Accident severity
Definition: Describes the severity of the road accident, based on the most severe injury of
any person involved.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Fatal: At least one person was killed immediately or died within 30 days as a result
of the road accident.
2. Serious/severe injury: At least one person was hospitalized for at least 24 hours
because of injuries sustained in the accident, while no one was killed.
3. Slight/minor injury: At least one of the participants of the accident was hospitalized
less than 24 hours or not hospitalized, while no participant was seriously injured or
killed.
Comments: Provides a quick reference to the accident severity, summarizing the data given
by the individual personal injury records of the accident. Facilitates analysis by accident
severity level.
3.2.1.3 Road data elements
The road related data elements describe the characteristics of the road and associated
infrastructure at the place and time of the accident.
R1. Type of roadway
Definition: Describes the type of road, whether the road has two directions of travel, and
whether the carriageway is physically divided. For accident occurring at junctions, where the
accident cannot be clearly allocated in one road, the road where the vehicle with priority was
moving is indicated.
Obligation: Mandatory
Data type: Numeric
34
Data values:
1. Motorway/freeway: Road with separate carriageways for traffic in two directions,
physically separated by a dividing strip not intended for traffic. Road has no crossings
at the same level with any other road, railway or tramway track, or footpath. Specially
sign-posted as a motorway and reserved for specified categories of motor vehicles.
2. Express road: Road with traffic in two directions, carriageways not normally
separated. Accessible only from interchanges or controlled junctions. Specially sign-
posted as an express road and reserved for specified categories of motor vehicles.
Stopping and parking on the running carriageway are prohibited.
3. Urban road, two-way: Road within the boundaries of a built-up area (an area with
sign-posted entries and exits). Single, undivided street with traffic in two directions,
relatively lower speeds (often up to 50 km/h), unrestricted traffic, with one or more
lanes which may or may not be marked.
4. Urban road, one-way: Road within the boundaries of a built-up area, with entries and
exits sign-posted as such. A single, undivided street with traffic in one direction,
relatively lower speeds (often up to 50 km/h).
5. Road outside a built-up area: Road outside the boundaries of a built-up area (an area
with sign-posted entries and exits).
6. Restricted road: A roadway with restricted access to public traffic. Includes cul-de-
sacs, driveways, lanes, private roads.
7. Other: Roadway of a type other than those listed above.
8. Unknown: Not known where the incident occurred.
Comments: Important for comparing accident rates of roads with similar design
characteristics, and for conducting comparative analyses between motorway and non-
motorway roads.
R2. Road functional class
Definition: Describes the character of service or function of the road where the first harmful
event took place. For accident occurring at junctions, where the accident cannot be clearly
allocated in one road, the road where the vehicle with priority was moving is indicated.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Principal arterial: Roads serving long distance and mainly interurban movements.
Includes motorways (urban or rural) and express roads. Principal arterials may cross
through urban areas, serving suburban movements. The traffic is characterized by
high speeds and full or partial access control (interchanges or junctions controlled by
traffic lights). Other roads leading to a principal arterial are connected to it through
side collector roads.
2. Secondary arterial: Arterial roads connected to principal arterials through
interchanges or traffic light-controlled junctions supporting and completing the urban
35
arterial network. Serving middle distance movements but not crossing through
neighborhoods. Full or partial access control is not mandatory.
3. Collector: Unlike arterials, collectors’ cross urban areas (neighborhoods) and collect
or distribute the traffic to/from local roads. Collectors also distribute traffic leading
to secondary or principal arterials.
4. Local: Roads used for direct access to the various land uses (private property,
commercial areas etc.). Low service speeds not designed to serve interstate or
suburban movements.
R3. Speed limit
Definition: The legal speed limit at the location of the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. nnn: The legal speed limit as provided by road signs or by the country’s traffic laws
for each road category, in kilometers per hour (km/h).
2. 999 (unknowns): The speed limit at the accident location is unknown.
Comments: For accident occurring at junctions, where the accident cannot be clearly
allocated in one road, the speed limit for the road where the vehicle with priority was moving
is indicated.
R4. Road obstacles
Definition: The presence of any person or object which obstructed the movement of the
vehicles on the road. Includes any animal standing or moving (either hit or not), and any
object not meant to be on the road. Does not include vehicles (parked or moving vehicles,
pedestrians) or obstacles on the side of the carriageway (e.g. poles, trees).
Obligation: Mandatory
Data type: Numeric
Data values:
1. Yes: Road obstacle(s) present at the accident site.
2. No: No road obstacle(s) present at the accident site.
3. Unknown: Unknown presence of any road obstacle(s) at the accident site.
R5. Road surface conditions
Definition: The condition of the road surface at the time and place of the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Dry: Dry and clean road surface.
2. Slippery: Slippery road surface due to existence of sand, gravel, mud, leaves, oil on
the road. Does not include snow, frost, ice or wet road surface.
3. Wet, damp: Wet road surface. Does not include flooding.
4. Flood: Still or moving water on the road.
5. Other: Other road surface conditions not mentioned above.
36
6. Unknown: The road surface conditions were unknown.
Comments: Important for identification of high wet-surface accident locations, for
engineering evaluation and prevention measures.
R6. Junction
Definition: Indicates whether the accident occurred at a junction (two or more roads
intersecting) and defines the type of the junction. In at-grade junctions all roads intersect at
the same level. In not-at-grade junctions’ roads do not intersect at the same level.
Obligation: Mandatory
Data type: Numeric
Data values:
1. At-grade, crossroad: Road intersection with four arms.
2. At-grade, roundabout: Circular road.
3. At-grade, T or staggered junction: Road intersection with three arms. Includes T
intersections and intersections with an acute angle.
4. At-grade, multiple junction: A junction with more than four arms (excluding
roundabouts).
5. At-grade, other: Other at-grade junction type not described above.
6. Not at grade: The junction includes roads that do not intersect at the same level.
7. Not at junction: The accident has occurred at a distance greater than 20 meters from
a junction.
8. Unknown: The accident location relative to a junction is unknown.
Comments: Accident occurring within 20 meters of a junction are considered as accident
accidents at a junction. Important for site-specific studies and identification of appropriate
engineering countermeasures.
R7. Traffic control at junction
Definition: Type of traffic control at the junction where accident occurred. Applies only to
accident accidents that occur at a junction.
Obligation: Mandatory if accident occurred at a junction (R6)
Data type: Numeric
Data values:
1. Authorized person: Police officer or traffic warden at intersection controls the traffic.
Applicable even if traffic signals or other junction control systems are present.
2. Stop signs: Priority is determined by stop sign(s).
3. Give-way sign or markings: Priority is determined by give-way sign(s) or markings.
4. Other traffic signs: Priority is determined by traffic sign(s) other than ‘stop’, ‘give
way’ or markings.
5. Automatic traffic signal (working): Priority is determined by a traffic signal that was
working at the time of the accident.
6. Automatic traffic signal (out of order): A traffic signal is present but out of order at
time of accident.
37
7. Uncontrolled: The junction is not controlled by an authorized person, traffic signs,
markings, automatic traffic signals or other means.
8. Other: The junction is controlled by means other than an authorized person, signs,
markings or automatic traffic signals.
Comments: If more than one value is applicable (e.g. traffic signs and automatic traffic
signals) record all that apply.
R8. Road curve
Definition: Indicates whether the accident occurred inside a curve, and what type of curve.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Tight curve: The accident occurred inside a road curve that was tight (based on the
judgment of the police officer).
2. Open curve: The accident occurred inside a road curve that was open (based on the
judgment of the police officer).
3. No curve: The accident did not occur inside a road curve.
4. Unknown: It is not defined whether the accident occurred inside a road curve.
Comments: Useful for identification and diagnosis of high-accident locations, and for
guiding changes to road design, speed limits, etc.
R9. Road segment grade
Definition: Indicates whether the accident occurred on a road segment with a steep gradient.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Yes: The accident occurred at a road segment with a high grade.
2. No: The accident did not occur at a road segment with a high grade.
3. Unknown: It is not defined whether the accident occurred at a road segment with a
high grade.
Comments: Useful for identification and diagnosis of high-accident locations, and for
guiding changes to road design, speed limits, etc.
3.2.1.4 Vehicle data elements
The vehicle data elements describe the characteristics and events of the vehicle(s) involved
in the accident.
V1. Vehicle number
Definition: Unique vehicle number assigned to identify each vehicle involved in the accident.
Obligation: Mandatory
Data type: Numeric, sequential two-digit number
Comments: Allows the vehicle record to be cross-referenced to the accident record and
person records.
38
V2. Vehicle type
Definition: The type of vehicle involved in the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Bicycle: Road vehicle with two or more wheels, generally propelled solely by the
energy of the person on the vehicle, in particular by means of a pedal system, lever
or handle.
2. Other non-motor vehicle: Another vehicle without engine not included in the list
above.
3. Two/three-wheel motor vehicle: Two or three-wheeled road motor vehicle (includes
mopeds, motorcycles, tricycles and all-terrain vehicles).
4. Passenger car: Road motor vehicle other than a two or three-wheeled vehicle,
intended for the carriage of passengers and designed to seat no more than nine (driver
included).
5. Bus/coach/trolley: Passenger-carrying vehicle, most commonly used for public
transport, inter-urban movements and tourist trips, seating more than nine persons.
Includes vehicles connected to electric conductors and which are not rail-borne.
6. Light goods vehicle (<3.5 t): Smaller (by weight) motor vehicle designed exclusively
or primarily for the transport of goods.
7. Heavy goods vehicle (≥3.5 t): Larger (by weight) motor vehicle designed exclusively
or primarily for the transport of goods.
8. Other motor vehicle: Other vehicle not powered by an engine and not included in the
two previous lists of values.
9. Unknown: The type of the vehicle is unknown or it was not stated.
Comments: Allows for analysis of accident risk by vehicle type and road user type (in
combination with Type of road user, P20). Important for evaluation of countermeasures
designed for specific vehicles or to protect specific road users.
V3. Vehicle makes
Definition: Indicate the make (distinctive name) assigned by motor vehicle manufacturer.
Obligation: Mandatory if the vehicle is a motorized vehicle. Not applicable to bicycles,
tricycles, rickshaws and animal-powered vehicles.
Data type: Character string. Alternatively, a list of motor vehicle makes can be composed,
with a code corresponding to each. Such a list allows for more consistent and reliable
recording, as well as for easier interpretation of the data.
Comments: Allows for accident analyses related to the various motor vehicle makes.
V4. Vehicle model
Definition: The code assigned by the manufacturer to denote a family of motor vehicles
(within a make) that have a degree of similarity in construction.
Obligation: Mandatory if the vehicle is a motorized vehicle. Not applicable to bicycles,
tricycles, rickshaws and animal-powered vehicles
39
Data type: Character string. Alternatively, a list of motor vehicle models can be composed,
with a code corresponding to each. Such a list allows for more consistent and reliable
recording, as well as for easier interpretation of the data.
Comments: Record the name of the model as referred to in the country in which the accident
occurred. Allows for accident analyses related to the various motor vehicle models.
V5. Vehicle model year
Definition: The year assigned to a motor vehicle by the manufacturer.
Obligation: Mandatory if the vehicle is a motorized vehicle. Not applicable to bicycles,
tricycles, rickshaws and animal-powered vehicles
Data type: Numeric (YYYY)
Comments: Can be obtained from vehicle registration. Important for use in identifying motor
vehicle model year for evaluation, research, and accident comparison purposes.
V6. Engine size
Definition: The size of the vehicle’s engine is recorded in cubic centimeters (cc).
Obligation: Mandatory, if vehicle is motorized. Not applicable to bicycles, tricycles,
rickshaws and animal-powered vehicles.
Data type: Numeric
Data values:
1. nnnn: Size of engine
2. 9999: Unknown engine size
Comments: Important for identifying the impact of motor vehicle power on accident risk.
V7. Vehicle special function
Definition: The type of special function being served by this vehicle regardless of whether
the function is marked on the vehicle.
Obligation: Mandatory, if vehicle is motorized. Not applicable to bicycles, tricycles,
rickshaws and animal-powered vehicles.
Data type: Numeric
Data values:
1. No special function: No special function of the vehicle.
2. Taxi: Licensed passenger car for hire with driver, without predetermined routes.
3. Vehicle used as bus: Passenger road motor vehicle used for the transport of people.
4. Police / military: Motor vehicle used for police / military purposes.
5. Emergency vehicle: Motor vehicle used for emergency purposes (includes
ambulances, fire service vehicles etc.).
6. Other: Other special functions, not mentioned above.
7. Unknown: It was not possible to record a special function.
Comments: Important to evaluate the accident involvement of vehicles used for special uses.
V8. Vehicle maneuvers
Definition: The controlled maneuver for this motor vehicle prior to the accident.
Obligation: Mandatory
40
Data type: Numeric
Data values:
1. Reversing: The vehicle was reversing.
2. Parked: Vehicle was parked and stationary.
3. Entering or leaving a parking position: The vehicle was entering or leaving a parking
position
4. Slowing or stopping: The vehicle was slowing or stopping
5. Moving off: The vehicle was still and started moving. Does not include vehicle
leaving or entering a parking position.
6. Waiting to turn: The vehicle was stationary, waiting to turn.
7. Turning: The vehicle was turning (includes U-turns).
8. Changing lane: The vehicle was changing lane.
9. Avoidance maneuvers: The vehicle changed its course in order to avoid an object on
the carriageway (including another vehicle or pedestrian).
10. Overtaking vehicle: The vehicle was overtaking another vehicle.
11. Straight forward / normal driving: The vehicle was moving ahead away from any
bend.
12. Other
13. Unknown
3.2.1.5 Person data elements
The person data elements describe the characteristics, actions, and consequences relating to
the people involved in the accident. These elements are to be completed for every person
injured in the accident, and also for the drivers of all vehicles (motorized and non-motorized)
involved in the accident.
P1. Person number
Definition: Number assigned to uniquely identify each person involved in the accident.
Obligation: Mandatory
Data type: Numeric (two-digit number, nn)
Comments: The persons related to the first (presumed liable) vehicle will be recorded first.
Within a specific vehicle, the driver will be recorded first, followed by the passengers. Allows
the person record to be cross-referenced to accident, road and vehicle records to establish a
unique linkage with the Accident ID (A1) and the Vehicle number (V1).
P2. Occupant’s vehicle number
Definition: The unique number assigned for this accident to the motor vehicle in which the
person was an occupant (V1).
Obligation: Mandatory
Data type: Numeric (two-digit number, nn)
Comments: Allows the person record to be cross-referenced to the vehicle records, linking
the person to the motor vehicle in which they were travelling.
41
P3. Pedestrian’s linked vehicle number
Definition: The unique number assigned for this accident to the motor vehicle which collided
with this person (V1). The vehicle number assigned under (V1) to the motor vehicle which
collided with this person.
Obligation: Mandatory
Data type: Numeric (two-digit number, nn, from V1)
Comments: Allows the person record to be cross-referenced to the vehicle records, linking
the person to the motor vehicle that struck them.
P4. Date of birth
Definition: Indicates the date of birth of the person involved in the accident.
Obligation: Mandatory
Data type: Numeric (date format – dd/mm/yyyy, 99/99/9999 if birth date unknown)
Comments: Allows calculation of person’s age. Important for analysis of accident risk by age
group, and assessing effectiveness of occupant protection systems by age group. Key variable
for linkage with records in other databases.
P5. Gender
Definition: Indicates the gender of the person involved in the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Male: On the basis of identification documents / personal ID number or determined
by the police.
2. Female: On the basis of identification documents / personal ID number or determined
by the police.
3. Unknown: Sex could not be determined (police unable to trace person, not specified).
Comments: Important for analysis of accident risk by sex. Important for evaluation of the
effect of sex of the person involved on occupant protection systems and motor vehicle design
characteristics.
P6. Type of road user
Definition: This variable indicates the role of each person at the time of the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Driver: Driver or operator of motorized or non-motorized vehicle. Includes cyclists,
persons pulling a rickshaw or riding an animal.
2. Passenger: Person riding on or in a vehicle, who is not the driver. Includes person in
the act of boarding, alighting from a vehicle or sitting/stranding.
3. Pedestrian: Person on foot, pushing or holding a bicycle, pram or a pushchair, leading
or herding an animal, riding a toy cycle, on roller skates, skateboard or skis. Excludes
persons in the act of boarding or alighting from a vehicle.
42
4. Other: Person involved in the accident who is not of any type listed above.
5. Unknown: It is not known what role the person played in the accident.
Comments: Allows for analysis of accident risk by road user type (in combination with
Vehicle type, V2). Important for evaluation of countermeasures designed to protect specific
road users.
P7. Seating position
Definition: The location of the person in the vehicle at the time of the accident.
Obligation: Mandatory for all vehicle occupants
Data type: Numeric
Subfield: Row
Data values:
1. Front
2. Rear
3. Not applicable (e.g. riding on motor vehicle exterior)
4. Other
5. Unknown
Subfield: Seat
Data values:
1. Left
2. Middle
3. Right
4. Not applicable (e.g. riding on motor vehicle exterior)
5. Other
6. Unknown
Comments: Important for full evaluation of occupant protection programmes.
P8. Injury severity
Definition: The injury severity level for a person involved in the accident.
Obligation: Mandatory
Data type: Numeric
Data values:
1. Fatal injury: Person was killed immediately or died within 30 days, as a result of the
accident.
2. Serious/severe injury: Person was hospitalized for at least 24 hours because of injuries
sustained in the accident.
3. Slight/minor injury: Person was injured and hospitalized for less than 24 hours or not
hospitalized.
4. No injury: Person was not injured.
5. Unknown: Injury severity was not recorded or is unknown.
43
Comment: Important for injury outcome analysis and evaluation and appropriate
classification of accident severity (PD1). Important element for linkage with records in other
databases.
P9. Safety equipment
Definition: Describes the use of occupant restraints, or helmet use by a motorcyclist or
bicyclist.
Obligation: Mandatory
Data type: Numeric
Subfield: Occupant restraints
Data values:
1. Seat-belt available, used
2. Seat-belt available, not used
3. Seat-belt not available
4. Child restraint system available, used
5. Child restraint system available, not used
6. Child restraint system not available
7. Not applicable: No occupant restraints could be used on the specific vehicle (e.g.
agricultural tractors).
8. Other restraints used
9. Unknown: Not known if occupant restraints were in use at the time of the accident.
10. No restraints used
Subfield: Helmet use
Data values:
1. Helmet worn
2. Helmet not worn
3. Not applicable (e.g. person was pedestrian or car occupant)
4. Unknown
Comments: Information on the availability and use of occupant restraint systems and helmets
is important for evaluating the effect of such safety equipment on injury outcomes.
P10. Pedestrian maneuvers
Definition: The action of the pedestrian immediately prior to the accident.
Obligation: Mandatory
Data type: Numeric
Data values
1. Crossing: The pedestrian was crossing the road.
2. Walking on the carriageway: The pedestrian was walking across the carriageway
facing or not facing traffic.
3. Standing on the carriageway: The pedestrian was on the carriageway and was
stationary (standing, sitting, lying etc.).
44
4. Not on the carriageway: The pedestrian was standing or moving on the sidewalk or
at any point beside the carriageway.
5. Other: The vehicle or the pedestrian was performing a maneuver not included in the
list of the previous values.
6. Unknown: The maneuvers performed by the vehicle or the pedestrian was not
recorded or it was unknown.
Comments: Provides useful information for the development of effective road design and
operation, education and enforcement measures to accommodate pedestrians.
P11. Alcohol use suspected
Definition: Law enforcement officer suspects that person involved in the accident has
consumed alcohol.
Obligation: Mandatory for all drivers of motorized vehicles, recommended for all non-
motorists (pedestrians and cyclists).
Data type: Numeric
Data values:
1. No
2. Yes
3. Not applicable (e.g. if person is not driver of motorized vehicle)
4. Unknown
P12. Alcohol test
Definition: Describes alcohol test status, type and result.
Obligation: Conditional (mandatory if alcohol use suspected, P25)
Data type: Numeric
Subfield: Test status
Data values:
1. Test not given
2. Test refused
3. Test given
4. Unknown if tested
Subfield: Test type
Data values:
1. Blood
2. Breath
3. Urine
4. Other
5. Test type unknown
Subfield: Test result
Data values
1. Value
2. Pending
45
3. Result unknown
Comments: Alcohol-related accidents are a major road safety problem. Information on
alcohol involvement in accident facilitates evaluation of programmes to reduce drink-
driving.
P13. Drug use
Definition: Indication of suspicion or evidence that person involved in the accident has
consumed illicit drugs.
Obligation: Mandatory for all drivers of motorized vehicles, recommended for all non-
motorists (pedestrians and cyclists).
Data type: Numeric
Data values:
1. No suspicion or evidence of drug use
2. Suspicion of drug use
3. Evidence of drug use (further subfields can specify test type and values)
4. Not applicable (e.g. if person is not driver of motorized vehicle)
5. Unknown
P14. Driving license issue date
Definition: Indicates the date (month and year) of issue of the person’s first driving license,
provisional or full, pertaining to the vehicle they were driving.
Obligation: Mandatory for all drivers of motorized vehicles
Data type: Numeric (MMYYYY)
Data values:
1. Value (MMYYYY)
2. Never issued a driving license
3. Date of issue of first license unknown
Comments: Allows calculation of number of years’ driving experience at the time of
accident.
3.2.2 Data collection and storage process
There are three primary methods by which accident data can be collected; police reports,
hospital reports and in-depth investigations.
3.2.2.1 Police reports
In most countries, the Police play a key role in the accident data collection process since they
are the first to arrive at the accident scene and record the needed data and are the last to
update the related data. The Police are also responsible for providing the authorities with the
collected data. Relevant authorities such as the police, ministries or governmental
departments are then responsible for maintaining the National accident data files and
publishing related statistics.
When called to an accident with casualties, the Police have to carry out an on-site
investigation and sometimes fill in an autopsy report as well as a part of the accident data
collection form. This form will be completed later at the police headquarters. When the 30-
46
days definition of fatalities is in place, the accident data forms have to be kept in the police
headquarters for at least one month and be finalized with the necessary updates for any killed
road users.
When the national road accident data are finalized, the Police are in charge of
forwarding the data to the body responsible for the national accident data file, e.g. the
National Statistical Office, the Ministry of Transport etc.
The main tool for accident data collection is the data collection form, hence the central
national authority responsible for the national accident file has to carry out the initial
development and the revisions later on, with the aim to cover not only the national needs but
also the international requirements.
The accident data collection form has to be coupled with clear instructions for filling
in, as well as for the data transmission process to the national data file. The national road
accident data form has to be revised regularly (at least once every ten years) in order to better
cope with the new needs of road accident analysis at national and international level, while
attention should be given to compatibility issues before and after the modifications.
The road accident data collection form should also include detailed information on the
accident type and conditions, the road infrastructure and the road and traffic environment.
Moreover, it should include detailed information on each vehicle involved in the accident
and on each road user (driver, passenger or pedestrian) affected by the accident.
Consequently, the national accident data collection form should be simple and self-
explaining in its structure. Moreover, the related instructions should be precise and detailed,
in order to provide clear and complete data definitions. It is also recommended that all
existing standardized international definitions of variables and values are adopted by the
national authorities when developing or revising their accident data collection forms.
Once the road accident data collection form is finalized by the Police, the form is
forwarded to the national authority responsible for maintaining the national road accident
data file. The necessary data quality control should then be undertaken within
Then, the data should be coded and introduced in the electronic national road accident
data file. Data coding includes the attribution of identification numbers to all accidents,
vehicles and persons involved, as well as the attribution of numerical codes to all data values.
It is also suggested to use different coding (i.e. groups of values) for the same variable, in
order to allow for different levels of detail to be directly available for the data users. For
example, it is common to code person age both in years and in age group classifications.
The structure of the national data file should be in accordance with the structure of the
accident data collection form. The use of sub-files, with each of them referring to the
accident, person and vehicle, would be efficient due to the hierarchical relationships of the
accident components. The different sub-files should be linked by means of the accident,
vehicle, road and person identification numbers, so that combined information on all accident
components can be easily retrieved. Thus, the national accident data file will include
disaggregate data for all accidents components, which can be retrieved by means of queries.
3.2.2.2 Hospital data
Data can be collected concerning road accident casualties who attend/are admitted to hospital
as a consequence of their accident. This provides the potential for the formation of a database
relating to Hospital Episodes.
47
For example, information on casualties admitted to hospital as in-patients in England
is contained in the Hospital Episodes Statistics (HES) database owned by the Information
Centre of the National Health Service (NHS). It is compiled by the Information Centre (IC)
from over 300 NHS Trusts in England. Casualties treated in Accident and Emergency
departments who are not subsequently admitted to a hospital are not included in the HES
database. However, all casualties admitted to a bed in a hospital in England should be
recorded in the data even if the admission did not require an overnight stay. International
standard diagnostic classifications are used in the health records (ICD-10). These include
transport accident codes which allow for the identification of road transport accident
casualties. More specifically, they allow the identification of road user type and casualty
class (e.g. casualty being a passenger of a motorcycle).
For this method, the hospital admissions records are based on periods of care (episodes)
under a particular consultant. So, a single patient may have more than one episode of care
arising from a single accident (e.g. if they transfer to another consultant). Therefore, some
data cleaning (de-duplication) needs to be carried out to identify records relating to the same
patient and same accident.
As with the Police data, clear guidelines for the collection and coding of variables to
be included in Hospital data are required. Identifiers should be put in place that allow
matching of hospital and police data in the event that both sources are collected within a
country. This enables a rich database to be developed that benefits from both the on-scene
report from the police and also the detailed injury outcome from the hospital.
3.2.2.3 In-depth accident investigations
In-depth accident data, sometimes termed microscopic data, is an ideal method to
identify and evaluate human factor issues related to real world accidents and potential Human
Machine Interface (HMI) issues faced by road users. The advantage of this data source is the
high level of detail known about each accident and how this can be related to a number of
outcomes. Microscopic data is usually collected by independent research teams with a strict
methodology collecting key variables pertaining to the accident, vehicle, road user, injury
data, interview information, road infrastructure and scene information, accident
reconstructions and accident causation analysis all of which is collected and analyzed by
experienced investigators.
The data collected by the in-depth collection activities is independent and transparent,
as opposed to the national reporting systems which are generally based on judicial
investigations, although these will be impartial investigations they will often be collected
with “vehicle to blame” in mind. In-depth accident data collected by the researchers is aimed
at the cause of the accident, not who was to blame (Hagstroem et al., 2010).
Accident investigations are undertaken in two ways; at the scene or retrospectively.
These are achieved by collecting data from accidents wither within minutes of their
occurrence, where a specialist investigation team attend the scene along with the emergency
services; or by retrospectively undertaking in-depth examinations of the vehicles and
recording their damage characteristics and assessing their crashworthiness.
The information gathered at the scene or retrospectively is enhanced with follow up
data including injury outcomes and causes for casualties who attend hospital and via
questionnaires sent to those involved in the accident along with any available witness
statements.
48
The data from in-depth accident investigations, whilst generally funded by a
governmental body, tend to be managed, stored and analyzed by research institutes
contracted by the government.
3.2.2.4 Representivity of accident data
When setting up accident data collection protocols at a country level, it is essential that
consideration be given to harmonization of these protocols across countries so that cross-
country comparative analyses can be made as robustly as possible. This has been considered
at a European level within several projects including DaCoTA where a common protocol for
European in-depth investigations was established (Atalar, Talbot, & Hill, 2012).
Once common national methods are in place, accident data from Police and Hospital
sources potentially provide the national picture in terms of the accident population and
resulting injury outcomes and therefore also have the potential to be fully representative of
the accident constellation.
For in-depth accident investigations, requiring specialist teams, sampling needs to be
taken into consideration in order to build a data base that is fit for the required analysis
purpose. To establish true representivity an ideal sampling plan would involve randomly
sampling accidents 24-7 all year round from regions that are nationally representative. This
however is not generally feasible due to practical and financial implications.
The DaCoTA project outlined the following principles for achieving a pan-European
representative accident sample for in-depth accidents (Hagstroem et al., 2010):
• Determine a sampling area which is representative of the national picture
• Within the sampling area, random sampling is considered a necessary precondition
to have broadly representative results.
• Stratification reduces the sample variance and still guarantees representativeness of
the sample.
• Multiple selection criteria (e.g. stratification according to different variables such as
road user type, accident severity) are possible provided the source of information is
reliable.
• Different strategies for sampling across regions / countries can be accommodated
provided they are undertaken consistently and transparently and as long as no (large)
biases in the sample are introduced.
3.3 Exposure data
Exposure indicators are considered indispensable in risk studies and international
comparisons. Multiple linkages of databases as well as systematic surveys of road user
behavior could facilitate the identification of relevant exposure data. However, for the
purposes of international comparisons and priority settings, efforts should be targeted in
defining exposure indicators as well as their compatibility to the accident data.
The exposure measures can be classified into two groups:
• Road traffic estimates: road length, vehicle kilometers and vehicle fleet.
• Road user at risk estimates: person kilometers, population, number of trips, time in
traffic and driver population.
49
Among these measures, vehicle fleet, driver population and road length are useful
alternative exposure measures in many countries worldwide, since the related data are
recorded systematically by most countries. However, the definitions used for the variables
and values are often not compatible.
Some basic requirements for the collection of such exposure measures are the following:
• Travel/mobility surveys for the collection of vehicles- or persons kilometers data
should be in the form required for accident risk analysis.
• Traffic counts systems have to be established on the national and main interurban
road network and at a later stage urban and rural areas to be included.
• A common vehicle classification should be considered by all countries.
• A common method for calculating vehicle-kilometers from the traffic counts should
be adopted.
The collection exposure data should be performed under a common framework in order
to obtain comparable indicators at international level. In this way, in the EU funded research
project SafetyNet the two-step methodology was developed for the EU countries. The
methodology includes:
1. harmonization of existing data and methods, including common transformation rules
for all countries and all exposure indicators, in order to improve their national
collection methods
2. collection of new harmonized data, including data collection at African level with
common definitions and methods.
The data needed for the estimation of the exposure indicators are the following:
• Road length data by road type, area type and region
• Vehicle fleet data by vehicle type and vehicle age
• Driver population data by driver age and gender
• Vehicle-kilometers by vehicle type, age, road type, area type
• Person-kilometers by person class, age and gender
Once these indicators have been harmonized and collected, additional data needs may be
tackled, such as:
• Vehicle fleet by engine type
• Driver population by nationality and experience
• Vehicle-kilometers by engine size
• Person-kilometers by nationality and experience
• Number of trips by person class, age, gender and vehicle type
• Time spent in traffic by person class, age, gender and vehicle type
3.3.1 Population
Population is a common exposure indicator used in road safety analyses due to the
availability of the related data. Three variables are useful when assessing accident risk at a
50
population level: person age, gender and nationality. In addition, population at regional level
would be important for calculating respective risks.
All variables and values (in particular person age, gender and nationality) included in
the population registers should have a straightforward meaning. Therefore, their definitions
and their compatibility should be assessed and used for any risk calculation in matching with
population-based road safety variables and values in the accident data base.
All countries have to collect population data in national registers and update them on a
regular basis by conducting nation-wide censuses. Considering that most censuses are carried
out on a regular basis (e.g. every 10 years), data for the intermediate years are estimations,
which are used for the annual updates of the registers.
Attention should be given to the character of population data. In general, international
databases provide average population data or population as of the 1st of January of every
year. To avoid misleading results, population data with the same characteristic should be
used.
However, for international comparisons risk calculations based on population data are
not sufficient, especially in the case of large differences of motorization level, traffic density
etc. among the countries. Therefore, additional exposure data have to be collected for risk
assessment.
3.3.2 Driver population
The best source for driver population data is usually the national driver licenses databases.
However, differences may exist among the countries concerning the registration of licensed
drivers in total or for specific vehicle types. In addition, errors or failure to update
systematically the register may lead to wrong estimations of the number of drivers. For
example, when individuals who have died or who are no longer licensed are not marked or
removed from the register there is an overestimation of the number of drivers.
Consequently, the following information should be available in the national registers on an
annual basis:
• the total number of active drivers’ licenses
• the number of drivers licenses by license group and by age group of the driver.
3.3.3 Road length
Road length data is a practical exposure variable for the estimation of traffic risk at the
network level. The variables selected have to be compatible with the respective accident data
concerning road. Thus, type of road, area type and region/municipality are regarded as useful
variables.
Information on road length by area type or region may be available in local authorities,
while for the main road network data may be available in national authorities. In order to
aggregate the existing information, the cooperation of several authorities responsible for the
operation and maintenance of road network is needed, while procedures such as national
questionnaires could be developed on that purpose.
If relevant data are not available, national authorities should carry out the required
activities for collecting this information. Road length data may be collected on-site, using
vehicles equipped with odometers, or with maps. In both cases, care must be taken in order
to adequately handle intersection areas and avoid double measuring their length.
51
3.3.4 Vehicle fleet
While the best estimation of exposure can be given by the number of vehicle-kilometers,
such data are not always available and are very expensive to collect. In the case that these
data are available, they are not always reliable. Therefore, the second-best exposure indicator
is considered to be the vehicle fleet, due to its correlation with the level of motorization.
Considering that the fatality risk is entirely different depending on the type of the
vehicle (e.g. bus, car, or bike) it is necessary to make the comparisons in the respect of
different vehicle categories. Consequently, the following information should be available in
the national registers on an annual basis:
• total number of registered vehicles
• number of vehicles by vehicle type and by age group of the vehicle.
3.3.5 Vehicle kilometers
As mentioned before, the number of vehicle-kilometers is probably the most appropriate
exposure indicator for the estimation of accident risk. Vehicle kilometers are a direct measure
of traffic volume and can be available in a significant level of disaggregation, i.e. time,
vehicle type, road type, driver characteristics etc.
However, in practice, the availability and the level of disaggregation of vehicle
kilometers varies significantly and is strongly dependent on the type and features of the
collection method used in each country. Moreover, the calculation of the exposure estimate
is not consistent throughout countries resulting in a low overall compatibility. Vehicle
kilometers are estimated by several methods, most of which include data collection by
surveys and traffic counts. Furthermore, estimations are also carried out by the use of
statistical models and combinations of methods.
In order to obtain a common and compatible risk exposure measurement unit, the
definition of the indicator should be uniform between all countries. In the Glossary of
Transport Statistics (Eurostat, 2003) a definition of vehicle kilometer is proposed, which
could form the basis for a common definition:
"Vehicle kilometer - Unit of measurement representing the movement of a road motor
vehicle over one kilometer. The distance to be considered is the distance actually run.
It includes movements of empty road motor vehicles. Units made up of a tractor and a
semi-trailer or a lorry and a trailer are counted as one vehicle”.
Vehicle kilometer data are most useful for traffic risk analyses related to the vehicle
and the road network. For the estimation of traffic risk at vehicle level, the vehicle type,
vehicle age, vehicle engine size and road type are the most important variables, while the
vehicle type, area type, road type and region variables are most important for the estimation
of traffic risk at network level.
3.3.6 Person kilometers
Person kilometers can be collected either by travel surveys or by traffic counts and occupancy
rate estimates. Travel surveys provide more detailed data than other methods. Moreover, data
on person kilometers for non-motorized road users (bicycles and pedestrians) as well as cross
tabulated data for age/gender groups of road users (both motorized and non-motorized) can
be obtained only through surveys.
52
Person-kilometer data estimated by surveys are more usable for the variables: person
class, person age and person gender and less usable for the vehicle type and the year.
However, data are collected through surveys based on all these indicators.
Travel surveys are currently the most promising method available in order to have
adequate data on person kilometers distributed by age/gender/road user. Thus, it is important
to design the surveys in ways that allow for relevant risk calculations to be made. It is
therefore recommended that travel surveys are conducted as follows:
• For risk exposure purposes travel surveys ought to be nationwide. Travel surveys in
particular areas are less suitable because it is difficult to know how representative the
area is, what the exact area covered is and it may be difficult to have precise
correspondence between exposure data and accident data.
• Travel surveys ought to have sub samples distributed over a whole year (for instance
sub samples every day) in order to account for seasonal travel variations.
• Travel surveys ought to include data also for professional drivers and travels
conducted as part of work in addition to private travels.
• Travel surveys based on person samples often lack data for children. A possible way
to obtain some data for children is to ask car drivers about age and gender of
passengers.
• It is important to distinguish between travel made in a road traffic environment and
travel made outside the road network. For pedestrians and cyclists this is particularly
relevant.
• In order to reduce the problems with inaccurate reporting of distances and time, one
should adopt tests of logic and reason to check answers.
• In addition to distance travelled one ought to try to register travel time as well.
3.4 Road Safe Performance Indicators
Safety performance indicators (SPIs) are measures (indicators), reflecting those operational
conditions of the road traffic system, which influence the system’s safety performance. SPIs
are aimed to serve as tools in assessing the current safety conditions of a road traffic system,
monitoring the progress, measuring impacts of various safety interventions and making
comparisons.
The performance indicators can be divided into four pillars - problem areas: road, vehicle,
road user and post-accident care. Indicative indicators on these four pillars consist of:
• road: number and length of road safety audits conducted, number of identified high
risk sites and related interventions
• vehicles: mean age of vehicle fleet, number of technical inspections
• road user: seat-belt use rates, helmet use rates, speeding, drink-driving and use of
mobile phone while driving
• post-accident care: number of staffs working on it, number of ambulances.
The present section presents the definitions of variables and values for producing national
SPIs in certain areas of the aforementioned pillars.
53
3.4.1 SPIs on drink-driving
Alcohol use by road users and especially by drivers of motor vehicles increases the road
accident risk considerably. Consequently, most countries ban the use of alcohol among
drivers, or set low legal limits for blood alcohol concentrations. Nevertheless, a high
proportion of fatal accidents involve drink-driving in most countries. Road safety policy
makers need information about the state of this problem in their countries.
A SPI reflecting the alcohol related road toll is the percentage of drivers under the
influence of alcohol.
Another more comparable indicator, which, however, seems to be out of line with the
basic idea of SPIs, is suggested in the SafetyNet project and is based on accident data. The
proposed SPI is the percentage of severe and fatal injuries resulting from road accidents
involving at least one active road user under the influence of alcohol.
In order to estimate the first indicator a sampling frame has to be defined, while for
the second one a national system has to be set up. Medically trained persons should take the
blood specimen and provide the respective results. It is also noted that amendments of the
road traffic law may be needed in countries where alcohol testing of drivers involved in fatal
accidents is not mandatory. The police should ensure that blood or breath samples are taken
from all drivers involved in road accidents and should report the results to the agency
responsible for national road accident statistics.
3.4.2 SPIs on the use of protection systems
The non-use of protection systems is associated with severe injuries and fatalities. Such
systems are the seat-belts for vehicle occupants, the helmets for riders of powered two-
wheelers and cyclists and the child restraint systems. The assessment of the use of protection
systems in traffic allows for identifying the magnitude of the problem and preventing fatal
injuries in road traffic.
The SPIs examined in this section are the following:
• wearing rates of seat belts, in front seats (passenger cars + vans /under 3.5 tons), in
rear seats (passenger cars + vans /under 3.5 tons), by children under 12 years old
(restraint systems use in passenger cars), and in front seats (HGV + coaches /above
3.5 tons)
• usage rates of safety helmets by cyclists, moped riders and motorcyclists.
The SPIs are estimated by conducting a national observational survey, where the
measurements should be classified by type of road, such as motorways, rural roads and urban
roads. The values for major road types are then aggregated into one indicator (of each type)
for the country. It is important that the assessment is conducted on a regular basis (preferably
annual).
3.4.3 SPIs on vehicles
The SPIs on vehicles are related to the level of protection afforded by the vehicles which
constitute the fleet in a country. When accidents occur, the potential of the vehicle itself to
prevent injuries can determine whether the outcome is a fatality or something less serious.
Thus, improvements in passive safety do not affect the occurrence of accidents, but help to
minimize the consequences when accidents happen. Unsafe operational conditions could be
defined as the presence within the fleet of a number of vehicles:
54
1. that will not protect the occupant well in a collision (accident worthiness)
2. with an increased capacity to inflict injury (compatibility).
The vehicles (passive safety) area differs from the other SPI areas, since the estimation
of the indicators is not based on surveys, but the necessary data are taken from national
databases. The minimum information which is required to produce some calculations of
vehicle age (as a proxy for vehicle accident worthiness) and fleet composition (as a measure
of compatibility), are total number of vehicles listed by:
• year of manufacture (or year of first registration)
• vehicle type (using definitions compatible with accident database).
3.5 Road Infrastructure Safety Management
Road Infrastructure Safety Management (RISM) refers to a set of procedures that support a
road authority in decision making related to road safety improvement of a road network.
These procedures are aimed at enhancing road safety at the different stages of a road
infrastructure life cycle (Figure 3-8). Some of them can be applied to existing infrastructures,
thus enabling a more reactive approach (e.g. by fixing the safety issues identified on the
infrastructure); while others are used in the early stages (i.e. planning and design) allowing a
more proactive approach (OECD/ITF, 2015).
Figure 3-8 Life cycle stages of a road infrastructure (OECD/ITF, 2015)
Several RISM procedures have been proposed in the last decades, some of them are
very popular (e.g. treatment of high-risk sites) and some are less known. In some cases, they
have similar characteristics. According to OECD/ITF (2015), the following are the most
consolidated RISM procedures:
• Road Safety Impact Assessment (RIA). A strategic comparative analysis of the impact
of a new road or a substantial modification to the existing network on the safety
performance of the road network. It is carried out at the initial planning stage before
1. Planning & Design
2. Construction & Pre-opening
3. NormalOperation
4. Maintenance& Renewal
5. Errorcorrection,
Hazardelimination
6. Major upgrading &
Renewal
55
the infrastructure project is approved. The purpose is to demonstrate, on a strategic
level, the implications on road safety of different planning alternatives of an
infrastructure project and these should play an important role when routes are
selected.
• Efficiency Assessment Tools (EAT). Budgets for transport in general and for road
safety in particular should be spent as optimally as possible. Efficiency assessment
tools (e.g. cost benefits analysis) determine the effects for society of an investment,
for instance of an investment in road safety, in order to prioritise investment
alternatives.
• Road Safety Audit (RSA). An independent detailed systematic and technical safety
check relating to the design characteristics of a road infrastructure project and
covering all stages, from planning to early operation, in order to identify and detail
unsafe features of a road infrastructure project.
• Network Operation (NO). This relates to daily management of the infrastructure of a
road network, with particular reference to maintaining road serviceability and safety.
• Road Infrastructure Safety Performance Indicators (SPIs). Safety performance
indicators (SPIs) are seen as any measurement that is causally related to crashes or
injuries and is used in addition to the figures of accidents or injuries, in order to
indicate safety performance or understand the process that leads to accidents. Road
Infrastructure Safety Performance Indicators aim to assess the safety hazards by
infrastructure layout and design (e.g. percentage of road network not satisfying safety
design standards).
• Network Safety Ranking (NSR). A method for identifying, analysing and classifying
parts of the existing road network according to their potential for safety development
and accident cost savings.
• Road Assessment Programmes (RAPs). These methods involve the collection of road
characteristics data which are then used to identify safety deficits or determine how
well the road environment protects the user from death or disabling injury when a
crash occurs.
• Road Safety Inspection (RSI). A preventive tool consisting of a regular, systematic,
on-site inspection of existing roads. The inspections cover the whole road network
and are carried out by trained safety expert teams. They result in in a formal report
on road hazards and safety issues found and which require a formal response by the
relevant road authority.
• High Risk Sites (HRS). A method to identify, analyse and rank sections of the road
network which have been in operation for more than three years and upon which a
large number of fatal accidents in proportion to the traffic flow have occurred.
• In-depth Investigation. In-depth Investigation is the acquisition of all relevant
information and the identification of one or several of the following: a) the cause (or
56
causes) of the accident; b) injuries, injury mechanisms and injury outcomes; c) how
the accident and injuries could have been prevented.
Why do we use RISM? Because as the time goes by, road infrastructure could change
in terms of performance and use. For instance, road conditions can change because of the
weather which worsen the status of pavements and so on. Road could also change in terms
of use: for instance, in terms of traffic volume (and we know road accidents change according
to traffic volume) or in terms of different road users.
Beyond the application to specific stages, other differences may appear when looking
at the type of road, the dimension of the tackled road safety problem (e.g. the entire road
network or a single road site) and the specific needs of the country using RISM procedures.
RISM procedures can be applied to every type of road, i.e. motorways, rural and urban roads.
However, some differences exist relating to “how” a procedure is carried out on a certain
type of road network, and the extent of the road network involved in the procedure (e.g. a
target site, a group of sites with similar characteristics or an area) (OECD/ITF, 2015).
Another aspect to take into account is the dimension of the road safety problem
examined – whether one is interested in studying a specific road section or intersection, a
road corridor or an entire road network. Some RISM procedures are applied to an entire road
network or to a part of it (e.g. Network Safety Ranking and High-Risk Sites rank road
sections) according to their safety level; therefore, they can be used only at network level (at
least two road sections). Other procedures, such as Road Safety Inspections, are applied at
section or intersection level. The use can be extended also to an entire road network, but
proceeding on a per-section basis (OECD/ITF, 2015). Table 3-1 outlines the road category
and extent of application for each RISM procedure.
Table 3-1 Context of application of RISM procedures (OECD/ITF, 2015)
Procedure Road Category Road Category
Road Safety Impact
Assessment
No specific road category Part of the road network potentially
influenced by a measure
Efficiency assessment
tools
No specific road category Part of the road network potentially
influenced by a measure
Road Safety Audit No specific road category A designed road infrastructure
Network Operation No specific road category,
however some practices are
difficult to perform on an urban
network
Generally part or an entire road network
managed by a road administration
Road Infrastructure
Safety Performance
Indicators
Usually performed on a rural and
motorway road network
An entire road network
Network Safety Ranking No specific road category Generally part or an entire road network
managed by a road administration
Road Assessment
Programs
Usually performed on a
rural/motorway road network
Part or an entire road network.
Road safety inspection No specific road category Generally part or all road elements belonging
to the same road network
57
High-Risk Sites No specific road category Generally part or an entire road network
managed by a road administration
In-depth Investigation No specific road category Limited to the area of intervention (e.g. 30
min from accident investigator’s base)
Another point to stress is the overlap of RISM procedures, meaning that in some cases,
two different procedures could lead to similar results or have some parts in common. This
may happen where some procedures have the same purpose, use the same tools or require
similar data (Figure 3-9). For example, Road Safety Audits (RSA), Road Assessment
Programmes (RAP), Road Safety Inspections (RSI), High-Risk Sites (HRS) and In-depth
Accident Investigations have in common a similar purpose: the identification of risk factors
related to road design or traffic control that may lead to accidents or make the accidents more
severe.
Figure 3-9 Data required and purposes associated to each procedure (OECD/ITF, 2015)
3.6 Road Infrastructure Safety Assessment Methodologies
A number of methodologies mostly based on the physical characteristics of a road have been
proposed over the last 15 years by researchers from around the world, especially from Italy
and New Zealand, so far to assess the safety performance of road infrastructures. As shown
in Table 3-2, most of the methodologies calculate a risk index, these have been concentrated
in segments of rural roads, and have a limited automated process in data collection, analysis,
and transmission of information.
58
Table 3-2 Summary of road safety infrastructure assessment methodologies
Literature reference Country Road Type Road
Element
Automated
Process Index
Montella (2005) Italy Rural roads Segments None Potential for Safety
Improvement Index
(PFI)
Cafiso et al. (2007) Italy Rural roads Segments None Safety Index (SI)
Appleton (2009), RISA New
Zealand
Rural roads Segments,
Intersection,
Road
network
None Personal Risk,
Collective Risk,
Network Risk
Number
iRAPa (2009) Worldwide Rural roads Segments Data
collection
Star Rating
Cafiso et al. (2011) Italy Low-volume
rural roads
Segments None Safety Index (SI)
NZTA (2013), High-
Risk Intersections
Guide
New
Zealand
Rural roads,
Urban streets
Intersections None Personal Risk,
Collective Risk
Brodie et al. (2013),
Urban KiwiRAP
New
Zealand
Urban streets Segments,
Intersections
Data
collection
Star Rating
Rosolino et al. (2014) Italy Rural roads Segments Real-time
information
to road users
on the risk
level in
relation to
their speed
Risk Index (RI)
Austroads (2014),
ANRAM
Australia Rural roads,
Urban streets
Segments,
Road
network
Software for
data analysis
Personal Risk,
Collective Risk
Zumrawi (2016) Sudan Rural roads Segments None Risk Factor Index
(RFI)
Hasmukhai (2016) India Urban streets Segments None Crash Risk Index
Chhanabhai et al.
(2017)
New
Zealand
Rural roads Segments None Infrastructure Risk
Rating (IRR)
Demasi et al. (2018) Italy Urban streets Segments None Branch Index Risk
(BIR), Section Index
Risk (SIR)
Note: aIncluding: EuroRAP, AusRAP, and usRAP
3.6.1 Road Infrastructure Assessment in Rural Roads
Recently, different methodologies for calculating a risk index on rural roads have been
proposed. In 2005, Montella (Montella, 2005) conducted a study to developing the potential
for a safety improvement index (PFI), the objective was to produce a technique to support
road safety inspections to quantify the safety gains that could be achieved by addressing the
problems identified in the review process. A systematic process to determine which road
features should be investigated and how each feature should be evaluated during the review
59
was described. The procedure addresses rural two-lane highways and does not take into
account junctions. From the process, the potential for a safety improvement index (PFI) was
calculated. Validation of the procedure was carried out by a comparison of the PFI values
with the expected collision frequency.
The PFI assessment is based on evaluation of safety items that have a known impact
on road safety. For each safety item, the relative increase in accident number and severity
was estimated. Safety reviewers, after a site investigation, by examination of videos recorded
during the inspection, identify the presence of individual features and measure the
approximate exposure length of each feature, dividing the road into homogeneous segments
of 200 m. Thus, ten general safety issues have been identified: alignment, cross-section,
markings, longitudinal rumble strips, pedestrian crosswalks, delineation, signs, pavement,
roadside, and accesses. PFI was assessed in 406 km of rural two-lane rolling highways in
Italy. Collision frequency was determined by application of a collision prediction model,
calibrated in the study network, and was refined by the application of the empirical Bayes
(EB) technique. Correlation between EB safety estimates and PFI values was highly
significant, with 93% of the variation in the estimated number of accidents explained by the
PFI value.
In Italy, in 2007 another study (Cafiso, Cava, & Montella, 2007) was carried out that
presented a methodological approach for the safety assessment of rural two-lane road
segments that use both analytical procedures referring to alignment design consistency
models and safety inspection processes. A safety index (SI) that quantitatively measures the
relative safety performance of a road segment was calculated. The SI measures the relative
safety performance of a road at intervals of 200 m. It does not consider junctions, and it refers
to two-lane rural highways. The SI was assessed in 30 segments of two-lane rural highways
in Italy. The following safety issues were assessed by using defined criteria: accesses, cross
section, delineation, markings, pavement, roadside, sight distance, and signs. The SI is
formulated by combining three components of risk: the exposure of road users to road
hazards, the probability of a vehicle’s being involved in an accident, and the resulting
consequences should an accident occur. This systematic and replicable procedure integrates
two different, complementary approaches—one based on design consistency evaluations and
the other on safety inspections—and makes it possible to address a wide variety of safety
issues effectively.
To test the procedure, comparisons were carried out between SI scores and the
empirical Bayes (EB) safety estimates. Validation of the procedure was carried out on a
sample of roads by a comparison of the risk rank obtained by using the SI and accident
history. Spearman’s rank correlation was used to determine the level of agreement between
the rankings obtained with the two techniques. The results from the Spearman’s rank–
correlation analysis validate the SI, indicating that the ranking from the SI scores and the EB
estimates agrees at the 99.9% level of significance with a correlation coefficient of 0.87. The
SI can be assessed whether accident data are available or not. If accident data are available
and are of good quality, the SI can be effectively used in conjunction with accident frequency
as ranking criteria. If accident data are not available or are unreliable, the SI can be used as
a proxy for accident data and becomes the only ranking criterion.
The New Zealand Transport Agency (NZTA) has developed a procedure called Road
Infrastructure Safety Assessment (RISA) (Appleton, 2009). RISA enables NZTA to monitor
a road controlling authority’s (RCA’s) performance over time with respect to road safety.
60
RISA provides the RCA with a tool to understand where the greatest road user benefits from
improved road safety infrastructure can be gained RISA has been developed as an evidence-
based tool following previous experience with safety auditing of existing roads. The main
results of a RISA are: the personal risk (risk to the individual driver), the Collective Risk
(risk to all road users), the Network Risk Number, the reduction in the Network Risk Number
for implementing network-wide treatments, the Safe Intersection Sight Distance Assessment,
the Intersection Safety Related Design Assessment, and the Intersection Safety Related
Maintenance Assessment.
RISA calculates the relative risk of each road assessed. The risk score is calculated per
km of road so that roads of unequal lengths may be compared. The risk scores are relative
risks and are called “Personal Risk”. A risk of 1.2 means that a person traveling on this road
has a 20% higher risk of a crash than when traveling on the benchmark road. As a general
rule low volume road have high risk relative to the benchmark road, and higher volume roads
have a relative risk closer to the benchmark road. Additionally, the traffic volume is
combined with the risk scores to create the “Collective Risk” i.e. the risk to all road users.
The Collective risk relates to crash numbers. RISA takes the Collective Risk Scores and data
on traffic volumes to scale up these results to the whole network and creates a Network Risk
Number. This is an abstract number. It relates to the number of crashes on the network.
Probably, the most known methodology is the international Road Assessment Program
(iRAP) (IRAP, 2009), the programme is the umbrella organisation for EuroRAP, AusRAP,
usRAP, and KiwiRAP. iRAP is based on four standardised protocols that together provide
consistent safety ratings of roads across borders. Nationally, they enable the identification of
the most dangerous roads, tracking performance over time, and therefore where the action is
appropriate. Internationally, they enable comparisons of risk within and between countries.
Standard protocols for iRAP are:
• Risk Mapping: based on real crash and traffic data, colour-coded maps show a
road's safety performance by measuring and mapping the rate at which people are
killed or seriously injured. Different maps can be produced depending on the target
audience.
• Performance Tracking: identifies whether fewer people are being killed or
seriously injured on individual routes or road networks over time, and importantly,
through consultation with road authorities, identifies the countermeasures that are
most effective.
• Star Rating: using drive-through inspections of routes in specially equipped
vehicles. Ratings show the likelihood of a crash occurring and how well the road
would protect against death or serious injury in the event of a crash.
• Safer Roads Investment Plans: Following road inspections and coding, in addition
to detailed reporting, a Safer Roads Investment Plan can be developed, considering
over 70 proven road improvement options.
iRAP consisting of a number of evaluation tools; among them, the most relevant to this
project is the Road Protection Score (RPS). The RPS module assigns a road infrastructure
safety level basing on how effectively the infrastructure prevents crashes and protects users
61
involved in crashes. Based on the calculated RPS the road section is classified according to
a five-level ranking (Star Rating).
iRAP methodology is the inspection of the road network in order to define the level of
safety inherent the road design: five-star roads (green) are the safest, and one-star (black) are
the least safe. Star Ratings can be completed without reference to detailed accident data,
which is often unavailable in low- and middle-income countries. Using specially equipped
vehicles, software and trained analysts, RAP inspections focus on more than 30 different road
design features that are known to influence the likelihood of a crash and its severity. These
features include intersection design, road cross-section and markings, roadside hazards,
footpaths, and bicycle lanes.
Two types of road inspections are available, drive-through inspections and video-based
inspections, with video-based inspections being the most common.
Drive-through inspections require inspectors to record road design data as they drive
along the road using a specialised data tablet. The process is technical and requires accredited
RAP inspectors. Drive-through inspections are typically used where the length of the road
network being surveyed is short or relatively simple (such as rural roads with no adjacent
development). The drive-through inspection equipment includes a video camera, touch-
sensitive laptop, and a GPS antenna. The inspections are followed by a period of data analysis
and quality checking.
Video-based inspections are undertaken in two stages. Firstly, a specially equipped
survey vehicle records images of the road as it travels along. The video is later viewed by
analysts, or coders, and assessed according to RAP protocols. The survey vehicle can record
digital images of the road (generally at intervals of 5-10 metres) using an array of cameras
aligned to pick up panoramic views of the road (forward, left-side and right-side). The main
forward view is calibrated to allow measurements such as lane width, shoulder width, and
distance to roadside hazards. The vehicles can drive along the road at almost normal speed
while collecting the information.
Following the completion of the video-based inspection, each relevant design feature
is measured and rated according to RAP protocols. The process involves streaming the video
images together to form a video of the road network. Coders then undertake desktop
inspections by conducting a virtual drive-through of the road network, at posted speed or on
a frame-by-frame basis, depending on the complexity of the road. The software used by the
coders enables accurate measurements of elements such as lane widths, shoulder widths, and
distance between the road edge and fixed hazards, such as trees or poles. To support the
process a detailed road inspection manual is available. At the completion of the rating
process, it is possible to produce a detailed condition report of the road that forms the basis
for Star Ratings and the Safer Roads Investment Plan. A colour coded map illustrating the
level of safety inherent the road design and features is produced and can be used to make
drivers aware of the risk of different roads or networks (OECD/ITF, 2015).
The Safety Index (SI) for low-volume roads (Cafiso, La Cava, & Montella, 2011)
measures the relative safety performance of a road segment. It does not take junctions into
account. The SI integrates two different approaches, one based on design consistency
evaluation and the other on safety inspections. The SI is formulated by combining three
components of risk: the exposure of road users to road hazards (exposure factor), the
probability of a vehicle’s being involved in an accident (accident frequency factor), and the
resulting consequences should an accident occur (accident severity factor). The SI was
62
assessed in 30 segments of low-volume roads in Italy and the identified safety issues were:
accesses, cross section, delineation, markings, pavement, roadside, sight distance, and signs.
The SI has two main practical applications. High-risk segments can be identified and ranked
by the SI score. Specific safety issues that contribute more to lack of safety are pointed out
in the RSI procedure in order to indicate more appropriate mass-action programs.
Within the Research Project “M2M – Mobile to Mobility: Information and
communication technology systems for road traffic safety”, another methodology was
developed for evaluating Road safety performance and proposed a new road network Risk
Index (RI) for info mobility (Rosolino et al., 2014). The RI is related to the risk deriving from
infrastructure’s features. In detail six different classes of events are identified: the number of
occurred accidents; density of intersections/accesses on the road section; road surface
anomalies; problems related to both horizontal and vertical road signs and deficiencies of
roadside and safety barriers. The research was focused on the possibility of giving real-time
information to road users about the risks associated with the specific travelled road segment,
using a multiplatform mobile application and a GPS system. The information is given to
drivers considering driver’s speeds (operating and average speeds) that are registered
continuously by the application.
The proposed methodology was validated by means of a pilot study composed of about
60 Km of a two-lane road in the district of Crotone (Calabria, Italy). The values of the Risk
Index estimated for some particular road segments were compared to the qualitative analysis
obtained by a Road Safety Inspection of the same test site. Results showed that the
methodology allows reaching a satisfactory matching between the two sets of data. However,
the authors recommended that more research is needed for a wider application of the
proposed method on several road types.
In 2008 the then Australian Transport Council (ATC), now known as the Standing
Council on Transport and Infrastructure (SCOTI), agreed to a number of measures that
should be progressed to further enhance Australia’s commitment to road safety. Within the
framework of these measures, The Australian National Risk Assessment Model (ANRAM)
was developed (Austroads, 2014). ANRAM helps road agencies identify fatal and serious
injury (severe) crash risk across all parts of the road network. ANRAM helps road agencies
manage this risk through the development of treatment programs aimed at reducing fatal and
serious injury crashes.
ANRAM uses risk assessment (iRAP v3 Beta 3 algorithms), crash prediction methods
and crashes history to identify road sections with a high risk of severe crashes. Risk
estimation is driven by relative safety performance of road infrastructure, traffic speed, flow,
and potential for vehicle conflicts. Severe crash history is used to supplement the predicted
results and account for road-user-related risk. Road sections may be ranked on the basis of
individual risk (ANRAM SRS) and collective risk (ANRAM FSI crashes). ANRAM includes
a Toolkit which enables scoping and comparison of proactive road safety treatment programs.
Such programs may range between high-cost Safe System Transformation works on highest-
risk parts of the road network to low-cost systemic improvements on the relatively safe parts
of the network. ANRAM enables estimation of economic benefits of proposed treatment
programs and benefit-cost ratios (BCRs).
The Risk Factor Index (RFI) (Mohamed Eltayeb Zumrawi, 2016) was established and
adapted to measure the safety hazards condition on the selected highways in Sudan. The RFI
is defined as a numerical indicator which rates the safety hazards condition of the existing
63
road. The RFI provides feedback on road safety performance for validation or improvement
of current road design and maintenance procedures. A numerical rating of the RFI ranges
from (0) to (10) with (0) being the lowest possible condition and (10) being the highest and
worst possible condition. A field survey of the current safety conditions was conducted on a
total of 3,350 Km of highways in Sudan. For each highway, the previously occurred
accidents, road surface problems, and inadequate traffic control facilities which directly
related to road safety were visually surveyed. The developed Risk Factor Index (RFI) was
considered in the analysis procedure to link each road with its information files containing
surveyed risk factors such as adverse geometric features conditions, poor road surface
conditions, accident history, problems related to traffic control devices, road lighting and
marking and roadside safety elements, and any other relevant risk factor.
The Infrastructure Risk Rating (IRR) (Chhanabhai, Beer, & Johnson, 2017), developed
in New Zealand, is a simplified-risk based road assessment methodology, based on fewer
features than other road risk tools. IRR scoring is based on the input of ten variables to
determine nine road features: Road stereotype; Carriageway width; Land Use; Access
Density; Speed; Alignment; Roadside Hazard Risk (Left and right side assessed separately
and averaged.); Intersection Density; Traffic Volume. Its outputs ratings over homogeneous
road lengths, and can use readily available imagery from Google Earth or Google Maps. The
study verified the applicability of the infrastructure risk rating (IRR) model on rural Victorian
roads by examining the relationship between the IRR model’s scores and historical crashes.
Similar previous analyses (Tate, 2015) showed that IRR correlates with the outputs of more
complicated road risk assessment programs such as KiwiRAP. In the same way, the IRR is a
good predictor of risk in relation to New Zealand roads.
Intersections are places on the road network where road users’ paths cross, increasing
the risk of a crash. Despite the relatively short time spent travelling through intersections on
most journeys, a high proportion of crashes occur at them (New Zeland Transport Agency -
NZTA, 2013). However, most of the methodologies have been developed for road segments
and very few for intersections.
The High-risk intersections guide follows in the footsteps of the High-risk rural roads
guide which the NZTA launched in September 2011. Both guides are a flagship Safer
Journeys 2020 initiative (New Zeland Transport Agency - NZTA, 2013). The High-risk
intersections guide introduces a new way to identify high-risk urban and rural intersections
and, using the Safe System approach, provides best practice guidance on how to identify,
prioritise and treat key road safety issues at high-risk intersections and the application of
proven countermeasures. High-risk intersections can be categorised using two types of risk
metrics: Collective and Personal Risk. Collective Risk is measured as the total number of
fatal and serious crashes or deaths and serious injury equivalents per intersection in a crash
period. Personal Risk is the risk of death or serious injuries to each vehicle entering the
intersection. The Personal Risk is calculated from the collective risk divided by a measure of
traffic volume.
The High-risk intersections guide also provides information on the most effective
measures to reduce casualties and severity by particular intersection form and control within
the overarching philosophy of a Safe System. There are four key treatment philosophies for
countermeasures for high-risk intersections. These are: safe system transformation
treatments, safer intersection treatments, safety management treatments, and safety
maintenance.
64
As has been shown previously, although the risk is calculated from the characteristics
of the infrastructure, each methodology considers different key factors. In this way, Table
3-3 shows the main attributes affecting road safety on rural roads.
Table 3-3 Summary of the main attributes affecting road safety on rural roads
Infrastructure/
Operational Element Specific Risk Factor
Mon
tell
a (
2005)
Cafi
so e
t al.
(2007)
Ap
ple
ton
(2009)
iRA
Pa (
2009)
Roso
lin
o e
t al.
(2014)
Au
stro
ad
s
(2014)
Zu
mra
wi
(2016)
Ch
han
ab
hai
et
al.
(2017)
Exposure Vehicle flow (AADT) x x x x x
Risks associated with traffic
composition (risk to VRUs only) x
Risks associated with the distribution of
traffic flow over arms at junctions x x x x
Speed Speed limit (general+motorcycle, truck) x
Operating speed x x x x x x
Mean speed x x
Road Surface Inadequate Friction x x x x x x x
Uneven surface x x x x x x x
Alignment - Road
Segments
Low Curve Radius x x x x x
Alignment deficiencies - High Grade x x x x
Poor sight distance – Horizontal curves x x x x x x x
Poor sight distance – Vertical curves x x x x x x
Cross-Section - Road
Segments
Number of lanes x x
Absence of paved shoulders x x
Lane width x x x x x x
Shoulder width x x x x x
Undivided Road - Median Type x x x x x x
Risks associated with safety barriers x x x x
Sight obstructions (Landscape,
Obstacles and Vegetation) x x x x x
Absence of guardrails or crash cushions x x x x
Absence of clear zone x
Missing passing lane x x
Missing climbing lane x
Traffic control – Road
segments
Absence of traffic signs x x x x x x
Absence of road markings x x x x x x
Absence of rumble strips x x x
Alignment and Traffic
Control - Junctions
Risk of different junction types x x x x
At-grade junction deficiencies -
Intersection quality x x x x x x
Density of intersection/lateral accesses x x x x x x x x
Uncontrolled rail-road crossing x x x
Poor junction readability - Absence of
road markings and crosswalks x x x
Road lighting Poor Visibility - Darkness (risk to
pedestrians only) x x
Poor Visibility - Darkness (risk to all) x x
65
Presence of
workzones
Roadworks x
Note: aIncluding: EuroRAP, AusRAP, and usRAP
3.6.2 Road Infrastructure Assessment in Urban Streets
As has been seen so far, most methodologies have been focused on rural roads and there are
few studies on road infrastructure safety assessment in the urban context. In 2012, the New
Zealand Transport Agency (NZTA) established a new KiwiRAP technical committee
charged with overseeing and directing the risk assessment process for roads in urban areas
and the development of an Urban KiwiRAP model. The existing KiwiRAP technical
committee was set up for the star rating model development for rural State Highways (Brodie,
Durdin, Fleet, Minnema, & Tate, 2013). KiwiRAP is part of an international family of Road
Assessment Programmes (RAP) under the umbrella of the International Road Assessment
Programme (iRAP). Urban KiwiRAP looks to apply road risk ratings to major urban
networks, use the Star Rating system and there are two components of the risk assessment
model; an intersection component and a corridor component. The risk assessment process for
intersections is defined by the High-Risk Intersections Guide and is applicable to
intersections in urban and rural environments.
Star Ratings measure and rate the safety of roads by considering a number of built-in
roads and roadside features. It involves a thorough visual assessment of many road and
roadside features including but not limited to: lane and shoulder width, horizontal alignment,
sight distance, and the location and nature of roadside objects. The visual assessment,
supplemented by high-speed data measurement, is carried out and recorded at 100m intervals
while the published Star Ratings are reported on segment lengths of at least 5km. In the same
way, Collective Risk and Personal Risk were established as risk metrics as part of KiwiRAP.
Collective Risk is based on the average annual number of fatal and serious crashes occurring
per kilometre of State Highway. Personal Risk is based on the average annual fatal and
serious injury crashes occurring per 100 million vehicle kilometers travelled.
The Crash Risk Index for urban roads (Hasmukhrai et al., 2016) was proposed in India
for evaluating urban road safety performance. Six factors were selected for the Crash Risk
Index calculation: the number of previously occurred accidents; density of
intersections/lateral accesses on the road section; road surface anomalies and irregularities;
problems related to horizontal road signs; problems related to vertical road signs; deficiency
of the roadside and safety barriers. A system architecture based on a user generated content
paradigm was built for evaluating the Crash Risk Index and informing drivers about the risk
associated to the road segment travelled, in order to make the transportation system safer and
more comfortable.
The proposed methodology was validated by means of a study on a road test-site in
Ahmadabad city, a sub-set of input parameters for the Crash Risk Index calculation was
selected. The values of the Crash Risk Index estimated for some particular road segments
were compared to the qualitative analysis obtained by a Road Safety Inspection of the same
test site. Results showed that the methodology allows reaching a satisfactory matching
between the two sets of data.
More than half of global road traffic deaths are amongst pedestrians, cyclists and
motorcyclists who are still too often neglected in road traffic system design in many countries
(World Health Organization, 2018). In this direction, an analytical methodology for the
assessment of the accident risk for Vulnerable Road Users (VRUs) (pedestrians, cyclists and
66
motorcyclists) in urban context was proposed (Demasi, Loprencipe, & Moretti, 2018). This
consists of a quantitative approach to assess the Branch Index Risk (BIR) and the Section
Index Risk (SIR) of existing urban roads considering their geometry, layout, users, and
traffic. The proposal relies on data collected during road safety inspections; therefore, it can
be implemented even when historical data about traffic volume or accidents are not available.
From the road inspections, the authors identified 9 categories of elements/defects of
infrastructure which could cause accidents: geometry; cross-section; private access;
pavement; lighting; road signs; intersection; urban furniture; and stopping.
The method depends on the assumed ranges of variables and risk classes, as well as on
the values attributed to the variables used for calculating the hazard index of examined
homogeneous road sections and branches. Therefore, both the Section Index Risk (SIR) and
the Branch Index Risk (BIR) depend on geometric, functional, physical, and environmental
defects or elements which are a potential source of road accidents. These factors are then
related to the involved vulnerable road users and to existing traffic flows to assess the current
levels of risk. The categorization of these values into six levels of risk allows the
identification of the most severe conditions and the prioritization of road safety works.
The proposed methodology was applied to multiple branches totaling 50 km together
in an Italian municipality in order to assess their BIR values. All the roads had the same
classification: two-lane urban roads with parking spaces and sidewalks on both sides and
their maximum allowable speed was 50 km/h.
3.7 Road Safety in Developing Countries: Evidence from SaferAfrica
project
Progress in reducing road traffic deaths over the last few years varies. Significantly between
the different regions and countries of the world. There continues to be a strong association
between the risk of a road traffic death and the income level of countries. With an average
rate of 27.5 deaths per 100,000 population, the risk is more than 3 times higher in low-income
countries than in high-income countries where the average rate is 8.3 deaths 100,000
population. As shown in Figure 3-10 the burden of road traffic deaths is disproportionately
high among low- and middle-income countries in relation to the size of their populations and
the number of motor vehicles in circulation. Although only 1% of the world's motor vehicles
are in low-income countries, 13% of deaths occur in these countries (World Health
Organization, 2018).
67
Figure 3-10 Proportion of population, road traffic deaths, and registered motor vehicles by country
income category (WHO, 2018)
According to the WHO (2018), countries in the Americas and Europe have the lowest
regional rates of 15.6 and 9.3 deaths per 100,000 people respectively. While Africa is the
worst performing continent in road safety (Figure 3-11). In the same way, in Africa there is
an observable difference between middle-income countries, which have a rate of death of
23.6 per 100,000 population and low-income countries, where the rate is 29.3 per 100,000
population.
Figure 3-11 Rates of road traffic death per 100,000 population by WHO regions:2013, 2016 (WHO,
2018)
In order to improve road safety performance in African countries, many barriers need
to be overcome. Among them stands the substantial lack of detailed knowledge on road
casualties in terms of their number as well as associated factors leading to road accidents or
affecting their consequences. There is a serious lack of road safety data in African countries,
and even when data are available (e.g. through the reports of WHO, International Road
Federation - IRF, etc.), little is known about data collection systems, data definitions, etc.
(Thomas et al., 2017)
In 2011 the Africa Road Safety Action Plan (ARSAP) established an Action Plan to
meet the objective of reducing road traffic crashes by 50% by the year 2020. Despite this
initiative, the situation worsens year after year. To contribute reverse this trend, the
SaferAfrica project, a joint effort of 16 partners from Africa and Europe, was launched in
2016. The SaferAfrica project was founded by the European Commission under the Horizon
2020 Mobility for Growth, carried out between October 2016 and September 2019. The
project aims at establishing a Dialogue Platform between Africa and Europe focused on road
safety and traffic management issues. It will represent a high-level body with the main
68
objective of providing recommendations to foster the adoption of specific initiatives,
properly funded.
The overall concept of SaferAfrica is depicted by a pyramid articulated in three levels,
shown in Figure 3-12. The top of the pyramid represents road safety and traffic management
actions oriented to the “Safe System approach”. The other two levels represent the Dialogue
Platform (DP). Of these two levels, the higher one is a decision‐making level, namely the
Institutional level, while the lower one constitutes the Technical level. These two levels are
closely interconnected to foster the appropriate match between African road safety policy
evolution, application, knowledge enhancement and institutional delivery capacity.
Figure 3-12 SaferAfrica overall concept (SaferAfrica, 2016)
The pyramid is based on the four building blocks, defined according to the priorities
highlighted by the Africa Road Safety Action Plan:
1. Road safety knowledge and data with the specific objective of setting up the African
Road Safety Observatory;
2. Road safety and traffic management capacity reviews;
3. Capacity building and training;
4. The sharing of good practices.
In order to assess the needs of stakeholders involved in road safety in terms of knowledge
and information tools and convey a clear view of current road safety practices followed in
Africa, two-fold surveys as well as existing road safety analysis documents were exploited.
The surveys consisted of a brief questionnaire in order to point out the current status in each
country in terms of basic road safety aspects and definitions, followed by an extensive one
where, besides other concerns, detailed demands and views of road safety stakeholders, not
necessarily directly involved in decision-making, in each examined African country were
69
recorded. Furthermore, existing road safety analysis documents were exploited; namely the
Global Status Report on Road Safety (WHO, 2015) and the IRF World Road Statistics 2016
(IRF, 2016) reports (Thomas et al., 2017).
This first survey addressed an initial approach to identify per country the current status
in terms of basic road safety management and data collection practices. Representatives from
20 African countries, mainly from the West, East and South regions of the African continent
took part in this survey. Most of the respondents had a significant experience in the field of
road safety (over 10 years), thus the information they provided is considered accurate and
reliable.
Experts from all countries stated emphatically the high importance of data and
knowledge to support road safety activities. This is a clear indication of the urgent need for
the improvement of data and information availability with regard to the improvement of road
safety in African countries.
The second survey included questions on road safety management and data collection
practices, road safety resources and basic road safety data developed appropriately to reflect
the conditions in Africa. This survey was filled-in by 29 stakeholders from 21 African
countries. The majority of the replies were received by governmental representatives.
All the information presented in the following section (2.1.1) is from Deliverable 4.1:
Survey results: road safety data, data collection systems and definitions (2017) of SaferAfrica
project.
3.7.1 Road safety data collection systems in Africa countries
3.7.1.1 General
The present section aims in clarifying the current status in terms of the existence, extent and
level of road safety data collection systems in African countries.
As an initial approach the existence of road safety databases and information at national
level in the examined countries was explored through question: "Do you use any national
databases/information sources? a. Road accident databases; b. travel/mobility survey results;
c. other exposure databases (e.g. vehicle fleet); d. other, please specify". Alternative answers
for each database/source: yes, no, don’t know).
From Figure 3-13 it can be seen that in most examined countries there are formal
systems in place for recording road accidents. Also, it is interesting to know that other
exposure databases are utilized in more than 50% of the countries. On the other hand, surveys
regarding travel or mobility demands seem not so widespread.
(a) (b) (c)
Notes: a: No feedback provided from Kenya, South Sudan, Senegal and Tunisia
70
b: No feedback provided from Benin, Kenya, Sierra Leone, South Africa, South Sudan, Senegal, Tanzania and
Tunisia.
c: No feedback provided from Gambia, Kenya, Sierra Leone, South Sudan, Senegal, Tanzania and Tunisia.
Figure 3-13 Existence and use of databases – information at national level
As a second approach, core road safety management concerns related to data collection
practices in the examined African countries, were addressed from the road safety monitoring
and evaluation points of view. The replies per country for these basic aspects, are shown in
Table 3-4. In the first column of Table 3-4, shortcuts of the questions on availability of road
safety management items are shown. The alternative answers were: yes, no, don't know.
Table 3-4 Basic aspects in monitoring and evaluation of road safety data collection practices in
African countries
Notes: √: Yes, Empty cell: No, N/A: No Answer, U/K: Unknown.
Experts revealed that sustainable and reliable systems (durable, funded and maintained)
to collect and manage data on road accidents, fatalities and injuries are available for a number
of African countries. On the other hand, sustainable in-depth accident investigations for road
safety purposes seem to be conducted for 8 out of 21 examined countries (Malawi,
Cameroon, D.R. of the Congo, Lesotho, Mali, Nigeria, Senegal and Togo). A national
observatory centralizing the data systems for road safety is available in almost 50% of the
responding countries. On the whole, the same countries also have a reporting procedure to
monitor road safety interventions in place. Last but not least, benchmarking is not really
utilized in most countries except for D.R. of the Congo, South Africa, Burkina Faso, Nigeria,
Sierra Leone and Tunisia.
3.7.1.2 Road accident data
As seen through Table 3-4, for 10 countries a national observatory is available for
centralizing the data systems for road safety. For these countries, different types of data
included in the national observatory were further specified through question: "Is there a
71
national Observatory centralizing the data systems for road safety? If yes, does it include data
on: accidents; fatalities or injuries; in-depth accident investigations; behavioural indicators;
exposure (traffic); violations or fines; driver licensing; vehicle registration; other data (please
specify)". Alternative answers were: yes, no, don't know.
Although in general such data vary, all 10 countries incorporate in their observatories
data on accidents, fatalities and injuries, 50% of them incorporate data regarding in-depth
accident investigations, and also 50%, data on behavioural indicators.
Monitoring road safety interventions through a reporting process is available for 8 of
the examined African countries (Table 3-4) (Question: "Has a reporting procedure been set
up to monitor the road safety interventions carried out in the country?"). Aiming to further
understand such practices in these countries, further questions were addressed and the results
are presented below.
The reporting of monitoring road safety interventions is mostly linked to intermediate
phases of the country’s national road safety programme as found in 4 out the 8 countries of
Table 3-4 (Question: "Is the reporting: periodical; linked to intermediate phases of the
RS programme?").
On the other hand, the most common areas of intervention to which the reporting
procedure applies are driver training, campaigns, enforcement and vehicle related measures
(Question: "Does reporting apply to all areas of intervention: Engineering measures on rural
roads; Planning and engineering interventions in urban areas; Enforcement operations;
Traffic education; RS campaigns; Driver training; Vehicle related measures; Others (please
specify").
Another interesting fact of the reporting process to monitor road safety interventions is
related to the level at which this is performed, which is mostly performed at regional/local
(60%) level and only in 3 countries at national level (covering ministries, government
agencies, etc.) as well (Questions: "Is reporting performed “horizontally” at the national level
(covering ministries and government agencies)?" and "Is reporting performed “vertically” to
cover activities at the regional and/or the local level?").
However, the information of this process is addressed mainly to the road safety lead
agency or the government itself (Question " Is the information addressed to?: the Lead
Agency; the high level inter-sectoral decision-making road safety institution; the technical
inter-sectoral road safety institution; the government; the Parliament?".
An additional but also important issue of concern is whether certain actions have been
taken based on the information collected through the reporting process and towards which
direction (Question: Has some action been taken on the basis of the outcome of this
information: limited changes in the action programme; allocation of funds or human
resources; training; others (please specify)) It was found that these actions in most cases
(75%) concern training as well as slight changes in the action programme, while allocation
of funds or human resources take place in less than 50% of these 8 countries.
Safety interventions need time to show results. However, it is important to check
whether such measures work as expected and do not generate undesired side-effects
(Question: "Does some "process evaluation" of safety interventions take place during the
implementation period of the programme (i.e. checking that measures work as expected and
do not generate undesired side-effects)?". It was found that such a process is undergoing in
approximately 35% of all the examined countries (Figure 3-14). Additional responses from
these 7 countries which provide further insight into this process are summarized below.
72
Notes: The number of respondents and the respective percentage per answer alternative are shown in the graph.
No feedback provided from South Sudan.
Figure 3-14 Existence of process evaluation for Safety interventions
It was found that in all 7 countries the evaluation for interventions addresses road safety
campaigns, in approximately 70% it addresses enforcement and vehicles and in around 50%
other areas (Question: "Is the evaluation for interventions addressing: all areas;
infrastructure; vehicles; enforcement; road safety campaigns; other areas (please specify)?").
The evaluation is performed using observations and/or field surveys or measurements
in 5 of the countries, whilst, for this task, safety performance indicators are utilized by 4
countries. (Question: "Does it involve: performance indicators; observations and/or field
surveys or measurements?").
Scientific expertise seems to be present in performing process evaluation in more than
50% of the countries (Question: "Are scientific expertise involved in performing process
evaluation?") while the evaluation results are available to all stakeholders in 70% of the
countries (Question: "Are the evaluation results available to all stakeholders?").
Finally, actions taken on the basis of the evaluation process results for most of these 7
countries involve both improvements of the implementation conditions and well as partial
changes in the action programme (Question: "Has some action been taken on the basis of the
outcome of this information such as: partial changes in the action programme; improvement
of implementation conditions?").
Furthermore, a process to assess the effects on accidents and injuries or socio-economic
costs of certain policy components seems to be available in 6 (29%) of the examined 21
countries (Question: "Has an evaluation process been planned to assess the effects on
accidents and injuries or socio-economic costs of some policy components (“product”
evaluation)?").
For these 6 countries the areas of interventions covered by the evaluation plan are
mainly enforcement and vehicle related measures, while infrastructure is slightly less covered
(Question: "Which areas of intervention are covered by the evaluation plan: infrastructure;
enforcement; vehicle related measures; others (please specify)?").
3.7.1.3 Risk exposure
The amount of travel in each country is one of the main determinants of road fatality risk.
However, traffic measurements are not systematically carried out in all countries. In general,
73
the lack of sufficient and reliable exposure data is still a major limitation of road safety
analyses and may significantly affect the potential for evidence-based policy making in the
African countries, regions and cities.
In terms of data collection systems, availability of exposure indicators was found in
the examined countries’ national observatories. As already discussed (Table 3-4), a national
observatory for centralizing the data systems for road safety seems to be available in 10
countries. From these 10 countries managing national observatories, approximately 50% (5
countries) seem to include exposure data in them.
3.7.1.4 Road safety performance indicators
In order to develop effective measures to reduce the number of accidents/injuries it is
necessary to understand the processes that lead to accidents. Safety Performance Indicators
(SPIs) can serve this purpose since by providing information, they serve as a link between
the casualties from road accidents and the measures to reduce them.
Road users’ behavioural aspects are a vital field of safety performance indicators. The
collection and management of such information are assessed through certain behavioural
indicators, such as speeding, drinking and driving, use of protection systems, distraction, etc.
Concerning data on behavioural indicators (Question: Are sustainable and reliable
systems in place to collect and manage data on behavioural indicators: vehicle speeds; safety
belt wearing rates; alcohol-impaired driving; others, please specify), a sustainable system for
their collection and management is in place for less than 50% of the 21 questioned countries.
For example, safety belt wearing rates are systematically collected and managed in fewer
countries (7 countries) compared to speeding and alcohol impaired driving (9 countries).
During the implementation period of a country’s national programme or policy, it is
very important to assess its safety performance (Question: Has a procedure been set up to
evaluate safety performances of the national programme or policy? If yes, are the
performances assessed on the basis of performance indicators; against national quantitative
targets?). Unfortunately, such a process is currently available in only 4 countries (19%),
where the safety performance is assessed based on national quantitative targets as well as on
performance indicators.